Amshaker commited on
Commit
647bcb6
·
verified ·
1 Parent(s): a1cde4b

Upload folder using huggingface_hub

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. outputs/2026-05-31/16-33-10/.hydra/config.yaml +235 -0
  2. outputs/2026-05-31/16-33-10/.hydra/hydra.yaml +232 -0
  3. outputs/2026-05-31/16-33-10/.hydra/overrides.yaml +74 -0
  4. outputs/2026-05-31/16-33-10/main.log +38 -0
  5. outputs/2026-05-31/16-33-11/.hydra/config.yaml +235 -0
  6. outputs/2026-05-31/16-33-11/.hydra/hydra.yaml +232 -0
  7. outputs/2026-05-31/16-33-11/.hydra/overrides.yaml +74 -0
  8. outputs/2026-05-31/16-33-11/main.log +39 -0
  9. outputs/2026-06-01/15-41-01/.hydra/config.yaml +234 -0
  10. outputs/2026-06-01/15-41-01/.hydra/hydra.yaml +232 -0
  11. outputs/2026-06-01/15-41-01/.hydra/overrides.yaml +74 -0
  12. outputs/2026-06-01/15-41-01/main.log +0 -0
  13. outputs/2026-06-01/15-41-06/.hydra/config.yaml +234 -0
  14. outputs/2026-06-01/15-41-06/.hydra/hydra.yaml +232 -0
  15. outputs/2026-06-01/15-41-06/.hydra/overrides.yaml +74 -0
  16. outputs/2026-06-01/15-41-06/main.log +0 -0
  17. outputs/2026-06-01/15-41-07/.hydra/config.yaml +234 -0
  18. outputs/2026-06-01/15-41-07/.hydra/hydra.yaml +232 -0
  19. outputs/2026-06-01/15-41-07/.hydra/overrides.yaml +74 -0
  20. outputs/2026-06-01/15-41-07/main.log +0 -0
  21. outputs/2026-06-01/15-41-08/.hydra/config.yaml +234 -0
  22. outputs/2026-06-01/15-41-08/.hydra/hydra.yaml +232 -0
  23. outputs/2026-06-01/15-41-08/.hydra/overrides.yaml +74 -0
  24. outputs/2026-06-01/15-41-08/main.log +5 -0
  25. outputs/2026-06-03/20-22-54/.hydra/config.yaml +234 -0
  26. outputs/2026-06-03/20-22-54/.hydra/hydra.yaml +232 -0
  27. outputs/2026-06-03/20-22-54/.hydra/overrides.yaml +74 -0
  28. outputs/2026-06-03/20-22-54/main.log +47 -0
  29. outputs/2026-06-03/20-22-55/.hydra/config.yaml +234 -0
  30. outputs/2026-06-03/20-22-55/.hydra/hydra.yaml +232 -0
  31. outputs/2026-06-03/20-22-55/.hydra/overrides.yaml +74 -0
  32. outputs/2026-06-03/20-22-55/main.log +10 -0
  33. outputs/2026-06-03/20-22-56/.hydra/config.yaml +234 -0
  34. outputs/2026-06-03/20-22-56/.hydra/hydra.yaml +232 -0
  35. outputs/2026-06-03/20-22-56/.hydra/overrides.yaml +74 -0
  36. outputs/2026-06-03/20-22-56/main.log +10 -0
  37. outputs/2026-06-04/06-22-27/.hydra/config.yaml +231 -0
  38. outputs/2026-06-04/06-22-27/.hydra/hydra.yaml +230 -0
  39. outputs/2026-06-04/06-22-27/.hydra/overrides.yaml +72 -0
  40. outputs/2026-06-04/06-22-27/main.log +0 -0
  41. outputs/2026-06-04/06-22-28/.hydra/config.yaml +231 -0
  42. outputs/2026-06-04/06-22-28/.hydra/hydra.yaml +230 -0
  43. outputs/2026-06-04/06-22-28/.hydra/overrides.yaml +72 -0
  44. outputs/2026-06-04/06-22-28/main.log +0 -0
  45. outputs/2026-06-04/06-24-27/.hydra/config.yaml +232 -0
  46. outputs/2026-06-04/06-24-27/.hydra/hydra.yaml +230 -0
  47. outputs/2026-06-04/06-24-27/.hydra/overrides.yaml +72 -0
  48. outputs/2026-06-04/06-24-27/main.log +77 -0
  49. outputs/2026-06-04/06-24-28/.hydra/config.yaml +232 -0
  50. outputs/2026-06-04/06-24-28/.hydra/hydra.yaml +230 -0
outputs/2026-05-31/16-33-10/.hydra/config.yaml ADDED
@@ -0,0 +1,235 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment:
2
+ debug: ${debug}
3
+ tasks:
4
+ - training
5
+ num_nodes: 1
6
+ training:
7
+ precision: 16-mixed
8
+ compile: false
9
+ lr: 2.0e-05
10
+ batch_size: 16
11
+ max_epochs: -1
12
+ max_steps: 350000
13
+ max_time: null
14
+ data:
15
+ num_workers: 4
16
+ shuffle: true
17
+ optim:
18
+ accumulate_grad_batches: 1
19
+ gradient_clip_val: 1.0
20
+ checkpointing:
21
+ every_n_train_steps: 2000
22
+ every_n_epochs: null
23
+ train_time_interval: null
24
+ save_last: true
25
+ enable_version_counter: false
26
+ validation:
27
+ precision: 16-mixed
28
+ compile: false
29
+ batch_size: 1
30
+ val_every_n_step: 2000
31
+ val_every_n_epoch: null
32
+ limit_batch: 8
33
+ inference_mode: true
34
+ data:
35
+ num_workers: 4
36
+ shuffle: false
37
+ test:
38
+ precision: 16-mixed
39
+ compile: false
40
+ batch_size: 1
41
+ limit_batch: 1
42
+ data:
43
+ num_workers: 4
44
+ shuffle: false
45
+ logging:
46
+ metrics: null
47
+ dataset:
48
+ debug: ${debug}
49
+ metadata: data/${dataset.name}/metadata.json
50
+ data_mean: 0.5
51
+ data_std: 0.5
52
+ save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
53
+ n_frames: 1000
54
+ context_length: 1
55
+ resolution: 128
56
+ observation_shape:
57
+ - 3
58
+ - ${dataset.resolution}
59
+ - ${dataset.resolution}
60
+ external_cond_dim: 0
61
+ validation_multiplier: 1
62
+ frame_skip: 1
63
+ action_cond_dim: 25
64
+ _name: video_minecraft_latent
65
+ precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
66
+ n_frames_valid: 700
67
+ customized_validation: true
68
+ memory_condition_length: 0
69
+ wo_updown: false
70
+ angle_range: 180
71
+ pos_range: 8
72
+ algorithm:
73
+ debug: ${debug}
74
+ lr: ${experiment.training.lr}
75
+ x_shape:
76
+ - 16
77
+ - 18
78
+ - 32
79
+ frame_stack: 1
80
+ frame_skip: ${dataset.frame_skip}
81
+ data_mean: ${dataset.data_mean}
82
+ data_std: ${dataset.data_std}
83
+ external_cond_dim: 0
84
+ context_frames: 100
85
+ weight_decay: 0.002
86
+ warmup_steps: 1000
87
+ optimizer_beta:
88
+ - 0.9
89
+ - 0.99
90
+ uncertainty_scale: 1
91
+ guidance_scale: 0.0
92
+ chunk_size: 1
93
+ scheduling_matrix: autoregressive
94
+ noise_level: random_all
95
+ causal: true
96
+ diffusion:
97
+ objective: pred_v
98
+ beta_schedule: sigmoid
99
+ schedule_fn_kwargs: {}
100
+ clip_noise: 20.0
101
+ use_snr: false
102
+ use_cum_snr: false
103
+ use_fused_snr: true
104
+ snr_clip: 5.0
105
+ cum_snr_decay: 0.96
106
+ timesteps: 1000
107
+ sampling_timesteps: 20
108
+ ddim_sampling_eta: 0.0
109
+ stabilization_level: 15
110
+ architecture:
111
+ network_size: 64
112
+ attn_heads: 4
113
+ attn_dim_head: 64
114
+ dim_mults:
115
+ - 1
116
+ - 2
117
+ - 4
118
+ - 8
119
+ resolution: ${dataset.resolution}
120
+ attn_resolutions:
121
+ - 16
122
+ - 32
123
+ - 64
124
+ - 128
125
+ use_init_temporal_attn: true
126
+ use_linear_attn: true
127
+ time_emb_type: rotary
128
+ n_frames: ${dataset.n_frames}
129
+ metadata: ${dataset.metadata}
130
+ action_cond_dim: 25
131
+ use_plucker: true
132
+ _name: dememwm_memory_dit
133
+ memory_token_cross_attention: true
134
+ memory_cross_attn_layers: null
135
+ memory_condition_length: 0
136
+ pose_cond_dim: 5
137
+ log_video: true
138
+ dememwm:
139
+ enabled: true
140
+ training_stage: stage_1
141
+ debug_force_all_streams: false
142
+ curriculum:
143
+ enabled: true
144
+ full_stage_start_step: 20000
145
+ freeze_vae: true
146
+ dit_freeze:
147
+ enabled: true
148
+ lr:
149
+ dememwm_modules: 4.0e-05
150
+ memory_adapters: 4.0e-05
151
+ full_dit: 1.0e-05
152
+ token_patch_size: 2
153
+ anchor:
154
+ enabled: true
155
+ anchor_indices:
156
+ - 0
157
+ - 1
158
+ - 2
159
+ - 3
160
+ allow_generated_as_anchor: false
161
+ diverse_selection: true
162
+ compress:
163
+ downsample_ratio: 3
164
+ dynamic:
165
+ enabled: true
166
+ exclude_latest_local_frames: 4
167
+ recent_frames: 4
168
+ conv_kernel_t: 3
169
+ conv_stride_t: 2
170
+ revisit:
171
+ enabled: true
172
+ deterministic_pose_retrieval: true
173
+ fov_overlap_threshold: 0.6
174
+ high_quality_fov_threshold: 0.7
175
+ plucker_weight: 0.1
176
+ max_frames: 2
177
+ fov_half_h: 52.5
178
+ fov_half_v: 37.5
179
+ fov_radius: 30.0
180
+ fov_yaw_samples: 25
181
+ fov_pitch_samples: 20
182
+ fov_depth_samples: 20
183
+ pose_preselect_topk: 64
184
+ plucker_grid_h: 4
185
+ plucker_grid_w: 4
186
+ plucker_focal_length: 0.35
187
+ compress:
188
+ downsample_ratio: 3
189
+ stage_policy:
190
+ noise_bucket_logging: true
191
+ eval_ablation:
192
+ enabled: false
193
+ branch: A_plus_D_plus_R_normal
194
+ generated_history_proxy:
195
+ enabled: true
196
+ start_step: 40000
197
+ ramp_steps: 40000
198
+ max_prob: 0.25
199
+ noise_std: 0.25
200
+ dropout_prob: 0.0
201
+ injection:
202
+ dit_hidden_size: 1024
203
+ anchor_gate: 1.0
204
+ dynamic_gate: 1.0
205
+ revisit_gate: 1.0
206
+ cache:
207
+ enabled: true
208
+ device: cpu
209
+ keep_raw_latents: all
210
+ keep_compressed_records: true
211
+ keep_prefix_anchors: true
212
+ eviction_policy: none
213
+ no_evict: true
214
+ clear_between_videos: true
215
+ max_records: null
216
+ on_capacity_exceeded: warn
217
+ checkpoint:
218
+ strict_dememwm_eval_load: true
219
+ n_tokens: 4
220
+ debug: false
221
+ wandb:
222
+ entity: turlin
223
+ project: DeMemWM
224
+ mode: offline
225
+ resume: null
226
+ load: null
227
+ auto_resume: true
228
+ resume_ckpt_path: null
229
+ name: train_dememwm_full_berzelius_8a100_bs16_global128_350k
230
+ output_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_full_berzelius_8a100_bs16_global128_350k/
231
+ customized_load: true
232
+ seperate_load: true
233
+ diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
234
+ vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
235
+ only_tune_memory: false
outputs/2026-05-31/16-33-10/.hydra/hydra.yaml ADDED
@@ -0,0 +1,232 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
4
+ sweep:
5
+ dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
6
+ subdir: ${hydra.job.num}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ formatters:
71
+ simple:
72
+ format: '[%(asctime)s][HYDRA] %(message)s'
73
+ handlers:
74
+ console:
75
+ class: logging.StreamHandler
76
+ formatter: simple
77
+ stream: ext://sys.stdout
78
+ root:
79
+ level: INFO
80
+ handlers:
81
+ - console
82
+ loggers:
83
+ logging_example:
84
+ level: DEBUG
85
+ disable_existing_loggers: false
86
+ job_logging:
87
+ version: 1
88
+ formatters:
89
+ simple:
90
+ format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
91
+ handlers:
92
+ console:
93
+ class: logging.StreamHandler
94
+ formatter: simple
95
+ stream: ext://sys.stdout
96
+ file:
97
+ class: logging.FileHandler
98
+ formatter: simple
99
+ filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
100
+ root:
101
+ level: INFO
102
+ handlers:
103
+ - console
104
+ - file
105
+ disable_existing_loggers: false
106
+ env: {}
107
+ mode: RUN
108
+ searchpath: []
109
+ callbacks: {}
110
+ output_subdir: .hydra
111
+ overrides:
112
+ hydra:
113
+ - hydra.mode=RUN
114
+ task:
115
+ - +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k
116
+ - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_full_berzelius_8a100_bs16_global128_350k/
117
+ - auto_resume=true
118
+ - experiment.tasks=[training]
119
+ - algorithm=dememwm_memory_dit
120
+ - +customized_load=true
121
+ - +seperate_load=true
122
+ - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
123
+ - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
124
+ - +only_tune_memory=false
125
+ - dataset=video_minecraft_latent
126
+ - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
127
+ - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
128
+ - dataset.n_frames=1000
129
+ - +dataset.n_frames_valid=700
130
+ - +dataset.customized_validation=true
131
+ - +dataset.memory_condition_length=0
132
+ - +dataset.wo_updown=false
133
+ - +dataset.angle_range=180
134
+ - +dataset.pos_range=8
135
+ - ++algorithm.n_tokens=4
136
+ - algorithm.x_shape=[16,18,32]
137
+ - ++algorithm.context_frames=100
138
+ - ++algorithm.log_video=true
139
+ - ++algorithm.diffusion.sampling_timesteps=20
140
+ - ++algorithm.dememwm.debug_force_all_streams=false
141
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
142
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
143
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
144
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
145
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
146
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
147
+ - ++algorithm.dememwm.anchor.enabled=true
148
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
149
+ - ++algorithm.dememwm.anchor.diverse_selection=true
150
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=3
151
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
152
+ - ++algorithm.dememwm.dynamic.enabled=true
153
+ - ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4
154
+ - ++algorithm.dememwm.dynamic.recent_frames=4
155
+ - ++algorithm.dememwm.revisit.enabled=true
156
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
157
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
158
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
159
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
160
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
161
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
162
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
163
+ - ++algorithm.dememwm.revisit.max_frames=2
164
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=3
165
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
166
+ - ++algorithm.dememwm.cache.enabled=true
167
+ - ++algorithm.dememwm.cache.device=cpu
168
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
169
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
170
+ - ++algorithm.dememwm.cache.eviction_policy=none
171
+ - ++algorithm.dememwm.cache.no_evict=true
172
+ - ++algorithm.dememwm.cache.clear_between_videos=true
173
+ - ++algorithm.dememwm.cache.max_records=null
174
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
175
+ - ++algorithm.dememwm.curriculum.enabled=true
176
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
177
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
178
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
179
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
180
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
181
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
182
+ - experiment.training.batch_size=16
183
+ - experiment.training.optim.accumulate_grad_batches=1
184
+ - experiment.validation.batch_size=1
185
+ - experiment.validation.limit_batch=8
186
+ - experiment.training.checkpointing.every_n_train_steps=2000
187
+ - experiment.validation.val_every_n_step=2000
188
+ - experiment.training.max_steps=350000
189
+ job:
190
+ name: main
191
+ chdir: null
192
+ 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
193
+ id: ???
194
+ num: ???
195
+ config_name: training
196
+ env_set: {}
197
+ env_copy: []
198
+ config:
199
+ override_dirname:
200
+ kv_sep: '='
201
+ item_sep: ','
202
+ exclude_keys: []
203
+ runtime:
204
+ version: 1.3.2
205
+ version_base: '1.3'
206
+ cwd: /proj/cvl/users/x_fahkh2/DeMemWM
207
+ config_sources:
208
+ - path: hydra.conf
209
+ schema: pkg
210
+ provider: hydra
211
+ - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations
212
+ schema: file
213
+ provider: main
214
+ - path: ''
215
+ schema: structured
216
+ provider: schema
217
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-05-31/16-33-10
218
+ choices:
219
+ cluster: null
220
+ algorithm: dememwm_memory_dit
221
+ dataset: video_minecraft_latent
222
+ experiment: exp_video
223
+ hydra/env: default
224
+ hydra/callbacks: null
225
+ hydra/job_logging: default
226
+ hydra/hydra_logging: default
227
+ hydra/hydra_help: default
228
+ hydra/help: default
229
+ hydra/sweeper: basic
230
+ hydra/launcher: basic
231
+ hydra/output: default
232
+ verbose: false
outputs/2026-05-31/16-33-10/.hydra/overrides.yaml ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ - +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k
2
+ - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_full_berzelius_8a100_bs16_global128_350k/
3
+ - auto_resume=true
4
+ - experiment.tasks=[training]
5
+ - algorithm=dememwm_memory_dit
6
+ - +customized_load=true
7
+ - +seperate_load=true
8
+ - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
9
+ - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
10
+ - +only_tune_memory=false
11
+ - dataset=video_minecraft_latent
12
+ - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
13
+ - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
14
+ - dataset.n_frames=1000
15
+ - +dataset.n_frames_valid=700
16
+ - +dataset.customized_validation=true
17
+ - +dataset.memory_condition_length=0
18
+ - +dataset.wo_updown=false
19
+ - +dataset.angle_range=180
20
+ - +dataset.pos_range=8
21
+ - ++algorithm.n_tokens=4
22
+ - algorithm.x_shape=[16,18,32]
23
+ - ++algorithm.context_frames=100
24
+ - ++algorithm.log_video=true
25
+ - ++algorithm.diffusion.sampling_timesteps=20
26
+ - ++algorithm.dememwm.debug_force_all_streams=false
27
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
28
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
29
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
30
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
31
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
32
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
33
+ - ++algorithm.dememwm.anchor.enabled=true
34
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
35
+ - ++algorithm.dememwm.anchor.diverse_selection=true
36
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=3
37
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
38
+ - ++algorithm.dememwm.dynamic.enabled=true
39
+ - ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4
40
+ - ++algorithm.dememwm.dynamic.recent_frames=4
41
+ - ++algorithm.dememwm.revisit.enabled=true
42
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
43
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
44
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
45
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
46
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
47
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
48
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
49
+ - ++algorithm.dememwm.revisit.max_frames=2
50
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=3
51
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
52
+ - ++algorithm.dememwm.cache.enabled=true
53
+ - ++algorithm.dememwm.cache.device=cpu
54
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
55
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
56
+ - ++algorithm.dememwm.cache.eviction_policy=none
57
+ - ++algorithm.dememwm.cache.no_evict=true
58
+ - ++algorithm.dememwm.cache.clear_between_videos=true
59
+ - ++algorithm.dememwm.cache.max_records=null
60
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
61
+ - ++algorithm.dememwm.curriculum.enabled=true
62
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
63
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
64
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
65
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
66
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
67
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
68
+ - experiment.training.batch_size=16
69
+ - experiment.training.optim.accumulate_grad_batches=1
70
+ - experiment.validation.batch_size=1
71
+ - experiment.validation.limit_batch=8
72
+ - experiment.training.checkpointing.every_n_train_steps=2000
73
+ - experiment.validation.val_every_n_step=2000
74
+ - experiment.training.max_steps=350000
outputs/2026-05-31/16-33-10/main.log ADDED
@@ -0,0 +1,38 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [2026-05-31 16:34:05,141][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8
2
+ [2026-05-31 16:34:13,901][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8
3
+ [2026-05-31 16:34:18,601][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8
4
+ [2026-05-31 16:34:18,897][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8
5
+ [2026-05-31 16:34:19,141][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8
6
+ [2026-05-31 16:34:19,288][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8
7
+ [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]
8
+ [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]
9
+ [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]
10
+ [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]
11
+ [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]
12
+ [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]
13
+ [2026-05-31 16:34:39,689][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
14
+ [2026-05-31 16:34:39,689][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
15
+ [2026-05-31 16:34:39,689][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
16
+ [2026-05-31 16:34:39,689][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
17
+ [2026-05-31 16:34:39,689][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
18
+ [2026-05-31 16:34:39,689][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
19
+ [2026-06-03 16:33:00,182][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15
20
+ [2026-06-03 16:33:00,183][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15
21
+ [2026-06-03 16:33:00,184][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
22
+ [2026-06-03 16:33:00,182][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15
23
+ [2026-06-03 16:33:00,184][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
24
+ [2026-06-03 16:33:00,184][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
25
+ [2026-06-03 16:33:00,183][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15
26
+ [2026-06-03 16:33:00,186][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
27
+ [2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15
28
+ [2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
29
+ [2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15
30
+ [2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
31
+ [2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15
32
+ [2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
33
+ [2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15
34
+ [2026-06-03 16:33:00,190][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
35
+ [2026-06-03 16:33:01,747][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15
36
+ [2026-06-03 16:33:01,748][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
37
+ [2026-06-03 16:33:01,748][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15
38
+ [2026-06-03 16:33:01,749][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
outputs/2026-05-31/16-33-11/.hydra/config.yaml ADDED
@@ -0,0 +1,235 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment:
2
+ debug: ${debug}
3
+ tasks:
4
+ - training
5
+ num_nodes: 1
6
+ training:
7
+ precision: 16-mixed
8
+ compile: false
9
+ lr: 2.0e-05
10
+ batch_size: 16
11
+ max_epochs: -1
12
+ max_steps: 350000
13
+ max_time: null
14
+ data:
15
+ num_workers: 4
16
+ shuffle: true
17
+ optim:
18
+ accumulate_grad_batches: 1
19
+ gradient_clip_val: 1.0
20
+ checkpointing:
21
+ every_n_train_steps: 2000
22
+ every_n_epochs: null
23
+ train_time_interval: null
24
+ save_last: true
25
+ enable_version_counter: false
26
+ validation:
27
+ precision: 16-mixed
28
+ compile: false
29
+ batch_size: 1
30
+ val_every_n_step: 2000
31
+ val_every_n_epoch: null
32
+ limit_batch: 8
33
+ inference_mode: true
34
+ data:
35
+ num_workers: 4
36
+ shuffle: false
37
+ test:
38
+ precision: 16-mixed
39
+ compile: false
40
+ batch_size: 1
41
+ limit_batch: 1
42
+ data:
43
+ num_workers: 4
44
+ shuffle: false
45
+ logging:
46
+ metrics: null
47
+ dataset:
48
+ debug: ${debug}
49
+ metadata: data/${dataset.name}/metadata.json
50
+ data_mean: 0.5
51
+ data_std: 0.5
52
+ save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
53
+ n_frames: 1000
54
+ context_length: 1
55
+ resolution: 128
56
+ observation_shape:
57
+ - 3
58
+ - ${dataset.resolution}
59
+ - ${dataset.resolution}
60
+ external_cond_dim: 0
61
+ validation_multiplier: 1
62
+ frame_skip: 1
63
+ action_cond_dim: 25
64
+ _name: video_minecraft_latent
65
+ precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
66
+ n_frames_valid: 700
67
+ customized_validation: true
68
+ memory_condition_length: 0
69
+ wo_updown: false
70
+ angle_range: 180
71
+ pos_range: 8
72
+ algorithm:
73
+ debug: ${debug}
74
+ lr: ${experiment.training.lr}
75
+ x_shape:
76
+ - 16
77
+ - 18
78
+ - 32
79
+ frame_stack: 1
80
+ frame_skip: ${dataset.frame_skip}
81
+ data_mean: ${dataset.data_mean}
82
+ data_std: ${dataset.data_std}
83
+ external_cond_dim: 0
84
+ context_frames: 100
85
+ weight_decay: 0.002
86
+ warmup_steps: 1000
87
+ optimizer_beta:
88
+ - 0.9
89
+ - 0.99
90
+ uncertainty_scale: 1
91
+ guidance_scale: 0.0
92
+ chunk_size: 1
93
+ scheduling_matrix: autoregressive
94
+ noise_level: random_all
95
+ causal: true
96
+ diffusion:
97
+ objective: pred_v
98
+ beta_schedule: sigmoid
99
+ schedule_fn_kwargs: {}
100
+ clip_noise: 20.0
101
+ use_snr: false
102
+ use_cum_snr: false
103
+ use_fused_snr: true
104
+ snr_clip: 5.0
105
+ cum_snr_decay: 0.96
106
+ timesteps: 1000
107
+ sampling_timesteps: 20
108
+ ddim_sampling_eta: 0.0
109
+ stabilization_level: 15
110
+ architecture:
111
+ network_size: 64
112
+ attn_heads: 4
113
+ attn_dim_head: 64
114
+ dim_mults:
115
+ - 1
116
+ - 2
117
+ - 4
118
+ - 8
119
+ resolution: ${dataset.resolution}
120
+ attn_resolutions:
121
+ - 16
122
+ - 32
123
+ - 64
124
+ - 128
125
+ use_init_temporal_attn: true
126
+ use_linear_attn: true
127
+ time_emb_type: rotary
128
+ n_frames: ${dataset.n_frames}
129
+ metadata: ${dataset.metadata}
130
+ action_cond_dim: 25
131
+ use_plucker: true
132
+ _name: dememwm_memory_dit
133
+ memory_token_cross_attention: true
134
+ memory_cross_attn_layers: null
135
+ memory_condition_length: 0
136
+ pose_cond_dim: 5
137
+ log_video: true
138
+ dememwm:
139
+ enabled: true
140
+ training_stage: stage_1
141
+ debug_force_all_streams: false
142
+ curriculum:
143
+ enabled: true
144
+ full_stage_start_step: 20000
145
+ freeze_vae: true
146
+ dit_freeze:
147
+ enabled: true
148
+ lr:
149
+ dememwm_modules: 4.0e-05
150
+ memory_adapters: 4.0e-05
151
+ full_dit: 1.0e-05
152
+ token_patch_size: 2
153
+ anchor:
154
+ enabled: true
155
+ anchor_indices:
156
+ - 0
157
+ - 1
158
+ - 2
159
+ - 3
160
+ allow_generated_as_anchor: false
161
+ diverse_selection: true
162
+ compress:
163
+ downsample_ratio: 3
164
+ dynamic:
165
+ enabled: true
166
+ exclude_latest_local_frames: 4
167
+ recent_frames: 4
168
+ conv_kernel_t: 3
169
+ conv_stride_t: 2
170
+ revisit:
171
+ enabled: true
172
+ deterministic_pose_retrieval: true
173
+ fov_overlap_threshold: 0.6
174
+ high_quality_fov_threshold: 0.7
175
+ plucker_weight: 0.1
176
+ max_frames: 2
177
+ fov_half_h: 52.5
178
+ fov_half_v: 37.5
179
+ fov_radius: 30.0
180
+ fov_yaw_samples: 25
181
+ fov_pitch_samples: 20
182
+ fov_depth_samples: 20
183
+ pose_preselect_topk: 64
184
+ plucker_grid_h: 4
185
+ plucker_grid_w: 4
186
+ plucker_focal_length: 0.35
187
+ compress:
188
+ downsample_ratio: 3
189
+ stage_policy:
190
+ noise_bucket_logging: true
191
+ eval_ablation:
192
+ enabled: false
193
+ branch: A_plus_D_plus_R_normal
194
+ generated_history_proxy:
195
+ enabled: true
196
+ start_step: 40000
197
+ ramp_steps: 40000
198
+ max_prob: 0.25
199
+ noise_std: 0.25
200
+ dropout_prob: 0.0
201
+ injection:
202
+ dit_hidden_size: 1024
203
+ anchor_gate: 1.0
204
+ dynamic_gate: 1.0
205
+ revisit_gate: 1.0
206
+ cache:
207
+ enabled: true
208
+ device: cpu
209
+ keep_raw_latents: all
210
+ keep_compressed_records: true
211
+ keep_prefix_anchors: true
212
+ eviction_policy: none
213
+ no_evict: true
214
+ clear_between_videos: true
215
+ max_records: null
216
+ on_capacity_exceeded: warn
217
+ checkpoint:
218
+ strict_dememwm_eval_load: true
219
+ n_tokens: 4
220
+ debug: false
221
+ wandb:
222
+ entity: turlin
223
+ project: DeMemWM
224
+ mode: offline
225
+ resume: null
226
+ load: null
227
+ auto_resume: true
228
+ resume_ckpt_path: null
229
+ name: train_dememwm_full_berzelius_8a100_bs16_global128_350k
230
+ output_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_full_berzelius_8a100_bs16_global128_350k/
231
+ customized_load: true
232
+ seperate_load: true
233
+ diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
234
+ vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
235
+ only_tune_memory: false
outputs/2026-05-31/16-33-11/.hydra/hydra.yaml ADDED
@@ -0,0 +1,232 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
4
+ sweep:
5
+ dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
6
+ subdir: ${hydra.job.num}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ formatters:
71
+ simple:
72
+ format: '[%(asctime)s][HYDRA] %(message)s'
73
+ handlers:
74
+ console:
75
+ class: logging.StreamHandler
76
+ formatter: simple
77
+ stream: ext://sys.stdout
78
+ root:
79
+ level: INFO
80
+ handlers:
81
+ - console
82
+ loggers:
83
+ logging_example:
84
+ level: DEBUG
85
+ disable_existing_loggers: false
86
+ job_logging:
87
+ version: 1
88
+ formatters:
89
+ simple:
90
+ format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
91
+ handlers:
92
+ console:
93
+ class: logging.StreamHandler
94
+ formatter: simple
95
+ stream: ext://sys.stdout
96
+ file:
97
+ class: logging.FileHandler
98
+ formatter: simple
99
+ filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
100
+ root:
101
+ level: INFO
102
+ handlers:
103
+ - console
104
+ - file
105
+ disable_existing_loggers: false
106
+ env: {}
107
+ mode: RUN
108
+ searchpath: []
109
+ callbacks: {}
110
+ output_subdir: .hydra
111
+ overrides:
112
+ hydra:
113
+ - hydra.mode=RUN
114
+ task:
115
+ - +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k
116
+ - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_full_berzelius_8a100_bs16_global128_350k/
117
+ - auto_resume=true
118
+ - experiment.tasks=[training]
119
+ - algorithm=dememwm_memory_dit
120
+ - +customized_load=true
121
+ - +seperate_load=true
122
+ - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
123
+ - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
124
+ - +only_tune_memory=false
125
+ - dataset=video_minecraft_latent
126
+ - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
127
+ - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
128
+ - dataset.n_frames=1000
129
+ - +dataset.n_frames_valid=700
130
+ - +dataset.customized_validation=true
131
+ - +dataset.memory_condition_length=0
132
+ - +dataset.wo_updown=false
133
+ - +dataset.angle_range=180
134
+ - +dataset.pos_range=8
135
+ - ++algorithm.n_tokens=4
136
+ - algorithm.x_shape=[16,18,32]
137
+ - ++algorithm.context_frames=100
138
+ - ++algorithm.log_video=true
139
+ - ++algorithm.diffusion.sampling_timesteps=20
140
+ - ++algorithm.dememwm.debug_force_all_streams=false
141
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
142
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
143
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
144
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
145
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
146
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
147
+ - ++algorithm.dememwm.anchor.enabled=true
148
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
149
+ - ++algorithm.dememwm.anchor.diverse_selection=true
150
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=3
151
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
152
+ - ++algorithm.dememwm.dynamic.enabled=true
153
+ - ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4
154
+ - ++algorithm.dememwm.dynamic.recent_frames=4
155
+ - ++algorithm.dememwm.revisit.enabled=true
156
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
157
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
158
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
159
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
160
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
161
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
162
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
163
+ - ++algorithm.dememwm.revisit.max_frames=2
164
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=3
165
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
166
+ - ++algorithm.dememwm.cache.enabled=true
167
+ - ++algorithm.dememwm.cache.device=cpu
168
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
169
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
170
+ - ++algorithm.dememwm.cache.eviction_policy=none
171
+ - ++algorithm.dememwm.cache.no_evict=true
172
+ - ++algorithm.dememwm.cache.clear_between_videos=true
173
+ - ++algorithm.dememwm.cache.max_records=null
174
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
175
+ - ++algorithm.dememwm.curriculum.enabled=true
176
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
177
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
178
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
179
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
180
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
181
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
182
+ - experiment.training.batch_size=16
183
+ - experiment.training.optim.accumulate_grad_batches=1
184
+ - experiment.validation.batch_size=1
185
+ - experiment.validation.limit_batch=8
186
+ - experiment.training.checkpointing.every_n_train_steps=2000
187
+ - experiment.validation.val_every_n_step=2000
188
+ - experiment.training.max_steps=350000
189
+ job:
190
+ name: main
191
+ chdir: null
192
+ 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
193
+ id: ???
194
+ num: ???
195
+ config_name: training
196
+ env_set: {}
197
+ env_copy: []
198
+ config:
199
+ override_dirname:
200
+ kv_sep: '='
201
+ item_sep: ','
202
+ exclude_keys: []
203
+ runtime:
204
+ version: 1.3.2
205
+ version_base: '1.3'
206
+ cwd: /proj/cvl/users/x_fahkh2/DeMemWM
207
+ config_sources:
208
+ - path: hydra.conf
209
+ schema: pkg
210
+ provider: hydra
211
+ - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations
212
+ schema: file
213
+ provider: main
214
+ - path: ''
215
+ schema: structured
216
+ provider: schema
217
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-05-31/16-33-11
218
+ choices:
219
+ cluster: null
220
+ algorithm: dememwm_memory_dit
221
+ dataset: video_minecraft_latent
222
+ experiment: exp_video
223
+ hydra/env: default
224
+ hydra/callbacks: null
225
+ hydra/job_logging: default
226
+ hydra/hydra_logging: default
227
+ hydra/hydra_help: default
228
+ hydra/help: default
229
+ hydra/sweeper: basic
230
+ hydra/launcher: basic
231
+ hydra/output: default
232
+ verbose: false
outputs/2026-05-31/16-33-11/.hydra/overrides.yaml ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ - +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k
2
+ - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_full_berzelius_8a100_bs16_global128_350k/
3
+ - auto_resume=true
4
+ - experiment.tasks=[training]
5
+ - algorithm=dememwm_memory_dit
6
+ - +customized_load=true
7
+ - +seperate_load=true
8
+ - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
9
+ - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
10
+ - +only_tune_memory=false
11
+ - dataset=video_minecraft_latent
12
+ - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
13
+ - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
14
+ - dataset.n_frames=1000
15
+ - +dataset.n_frames_valid=700
16
+ - +dataset.customized_validation=true
17
+ - +dataset.memory_condition_length=0
18
+ - +dataset.wo_updown=false
19
+ - +dataset.angle_range=180
20
+ - +dataset.pos_range=8
21
+ - ++algorithm.n_tokens=4
22
+ - algorithm.x_shape=[16,18,32]
23
+ - ++algorithm.context_frames=100
24
+ - ++algorithm.log_video=true
25
+ - ++algorithm.diffusion.sampling_timesteps=20
26
+ - ++algorithm.dememwm.debug_force_all_streams=false
27
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
28
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
29
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
30
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
31
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
32
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
33
+ - ++algorithm.dememwm.anchor.enabled=true
34
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
35
+ - ++algorithm.dememwm.anchor.diverse_selection=true
36
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=3
37
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
38
+ - ++algorithm.dememwm.dynamic.enabled=true
39
+ - ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4
40
+ - ++algorithm.dememwm.dynamic.recent_frames=4
41
+ - ++algorithm.dememwm.revisit.enabled=true
42
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
43
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
44
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
45
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
46
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
47
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
48
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
49
+ - ++algorithm.dememwm.revisit.max_frames=2
50
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=3
51
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
52
+ - ++algorithm.dememwm.cache.enabled=true
53
+ - ++algorithm.dememwm.cache.device=cpu
54
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
55
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
56
+ - ++algorithm.dememwm.cache.eviction_policy=none
57
+ - ++algorithm.dememwm.cache.no_evict=true
58
+ - ++algorithm.dememwm.cache.clear_between_videos=true
59
+ - ++algorithm.dememwm.cache.max_records=null
60
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
61
+ - ++algorithm.dememwm.curriculum.enabled=true
62
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
63
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
64
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
65
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
66
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
67
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
68
+ - experiment.training.batch_size=16
69
+ - experiment.training.optim.accumulate_grad_batches=1
70
+ - experiment.validation.batch_size=1
71
+ - experiment.validation.limit_batch=8
72
+ - experiment.training.checkpointing.every_n_train_steps=2000
73
+ - experiment.validation.val_every_n_step=2000
74
+ - experiment.training.max_steps=350000
outputs/2026-05-31/16-33-11/main.log ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [2026-05-31 16:33:58,507][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP)
2
+ [2026-05-31 16:33:58,787][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True
3
+ [2026-05-31 16:33:58,787][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores
4
+ [2026-05-31 16:33:58,787][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs
5
+ [2026-05-31 16:33:58,787][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs
6
+ [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=''
7
+ [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=''
8
+ [2026-05-31 16:34:05,142][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8
9
+ [2026-05-31 16:34:14,613][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8
10
+ [2026-05-31 16:34:20,380][pytorch_lightning.utilities.rank_zero][INFO] - ----------------------------------------------------------------------------------------------------
11
+ distributed_backend=nccl
12
+ All distributed processes registered. Starting with 8 processes
13
+ ----------------------------------------------------------------------------------------------------
14
+
15
+ [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]
16
+ [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]
17
+ [2026-05-31 16:34:36,691][lightning.pytorch.callbacks.model_summary][INFO] -
18
+ | Name | Type | Params
19
+ -------------------------------------------------------------------------------------
20
+ 0 | diffusion_model | Diffusion | 910 M
21
+ 1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 199 K
22
+ 2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 591 K
23
+ 3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 591 K
24
+ 4 | dememwm_revisit_gate | RevisitRawGate | 4
25
+ 5 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M
26
+ 6 | vae | AutoencoderKL | 229 M
27
+ -------------------------------------------------------------------------------------
28
+ 911 M Trainable params
29
+ 231 M Non-trainable params
30
+ 1.1 B Total params
31
+ 4,573.841 Total estimated model params size (MB)
32
+ [2026-05-31 16:34:39,689][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
33
+ [2026-05-31 16:34:39,689][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
34
+ [2026-06-03 16:33:00,183][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15
35
+ [2026-06-03 16:33:00,184][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
36
+ [2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15
37
+ [2026-06-03 16:33:00,189][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
38
+ [2026-06-03 16:33:01,749][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15
39
+ [2026-06-03 16:33:01,750][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
outputs/2026-06-01/15-41-01/.hydra/config.yaml ADDED
@@ -0,0 +1,234 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment:
2
+ debug: ${debug}
3
+ tasks:
4
+ - training
5
+ num_nodes: 1
6
+ training:
7
+ precision: 16-mixed
8
+ compile: false
9
+ lr: 2.0e-05
10
+ batch_size: 16
11
+ max_epochs: -1
12
+ max_steps: 350000
13
+ max_time: null
14
+ data:
15
+ num_workers: 4
16
+ shuffle: true
17
+ optim:
18
+ accumulate_grad_batches: 1
19
+ gradient_clip_val: 1.0
20
+ checkpointing:
21
+ every_n_train_steps: 2000
22
+ every_n_epochs: null
23
+ train_time_interval: null
24
+ save_last: true
25
+ enable_version_counter: false
26
+ validation:
27
+ precision: 16-mixed
28
+ compile: false
29
+ batch_size: 1
30
+ val_every_n_step: 2000
31
+ val_every_n_epoch: null
32
+ limit_batch: 2
33
+ inference_mode: true
34
+ data:
35
+ num_workers: 4
36
+ shuffle: false
37
+ test:
38
+ precision: 16-mixed
39
+ compile: false
40
+ batch_size: 1
41
+ limit_batch: 1
42
+ data:
43
+ num_workers: 4
44
+ shuffle: false
45
+ logging:
46
+ metrics: null
47
+ dataset:
48
+ debug: ${debug}
49
+ metadata: data/${dataset.name}/metadata.json
50
+ data_mean: 0.5
51
+ data_std: 0.5
52
+ save_dir: /share_1/users/bonan_ding/worldmem_data/minecraft
53
+ n_frames: 1000
54
+ context_length: 1
55
+ resolution: 128
56
+ observation_shape:
57
+ - 3
58
+ - ${dataset.resolution}
59
+ - ${dataset.resolution}
60
+ external_cond_dim: 0
61
+ validation_multiplier: 1
62
+ frame_skip: 1
63
+ action_cond_dim: 25
64
+ _name: video_minecraft_latent
65
+ precomputed_feature_dir: /share_1/users/bonan_ding/worldmem_data/minecraft/vae_features
66
+ n_frames_valid: 1100
67
+ customized_validation: true
68
+ wo_updown: false
69
+ angle_range: 180
70
+ pos_range: 8
71
+ algorithm:
72
+ debug: ${debug}
73
+ lr: ${experiment.training.lr}
74
+ x_shape:
75
+ - 16
76
+ - 18
77
+ - 32
78
+ frame_stack: 1
79
+ frame_skip: ${dataset.frame_skip}
80
+ data_mean: ${dataset.data_mean}
81
+ data_std: ${dataset.data_std}
82
+ external_cond_dim: 0
83
+ context_frames: 100
84
+ weight_decay: 0.002
85
+ warmup_steps: 1000
86
+ optimizer_beta:
87
+ - 0.9
88
+ - 0.99
89
+ uncertainty_scale: 1
90
+ guidance_scale: 0.0
91
+ chunk_size: 1
92
+ scheduling_matrix: autoregressive
93
+ noise_level: random_all
94
+ causal: true
95
+ diffusion:
96
+ objective: pred_v
97
+ beta_schedule: sigmoid
98
+ schedule_fn_kwargs: {}
99
+ clip_noise: 20.0
100
+ use_snr: false
101
+ use_cum_snr: false
102
+ use_fused_snr: true
103
+ snr_clip: 5.0
104
+ cum_snr_decay: 0.96
105
+ timesteps: 1000
106
+ sampling_timesteps: 20
107
+ ddim_sampling_eta: 0.0
108
+ stabilization_level: 15
109
+ architecture:
110
+ network_size: 64
111
+ attn_heads: 4
112
+ attn_dim_head: 64
113
+ dim_mults:
114
+ - 1
115
+ - 2
116
+ - 4
117
+ - 8
118
+ resolution: ${dataset.resolution}
119
+ attn_resolutions:
120
+ - 16
121
+ - 32
122
+ - 64
123
+ - 128
124
+ use_init_temporal_attn: true
125
+ use_linear_attn: true
126
+ time_emb_type: rotary
127
+ n_frames: ${dataset.n_frames}
128
+ metadata: ${dataset.metadata}
129
+ action_cond_dim: 25
130
+ use_plucker: true
131
+ _name: dememwm_memory_dit
132
+ memory_token_cross_attention: true
133
+ memory_cross_attn_layers:
134
+ - 3
135
+ - 7
136
+ - 11
137
+ - 15
138
+ pose_cond_dim: 5
139
+ log_video: true
140
+ dememwm:
141
+ enabled: true
142
+ training_stage: stage_1
143
+ debug_force_all_streams: false
144
+ curriculum:
145
+ enabled: true
146
+ full_stage_start_step: 20000
147
+ freeze_vae: true
148
+ dit_freeze:
149
+ enabled: true
150
+ lr:
151
+ dememwm_modules: 4.0e-05
152
+ memory_adapters: 4.0e-05
153
+ full_dit: 1.0e-05
154
+ spatial_kernel_size: 3
155
+ anchor:
156
+ enabled: true
157
+ anchor_indices:
158
+ - 0
159
+ - 1
160
+ - 2
161
+ - 3
162
+ allow_generated_as_anchor: false
163
+ diverse_selection: true
164
+ compress:
165
+ downsample_ratio: 2
166
+ dynamic:
167
+ enabled: true
168
+ recent_frames: 4
169
+ conv_kernel_t: 3
170
+ conv_stride_t: 2
171
+ compress:
172
+ downsample_ratio: 2
173
+ revisit:
174
+ enabled: true
175
+ deterministic_pose_retrieval: true
176
+ fov_overlap_threshold: 0.6
177
+ plucker_weight: 0.1
178
+ max_frames: 2
179
+ fov_half_h: 52.5
180
+ fov_half_v: 37.5
181
+ fov_radius: 30.0
182
+ fov_yaw_samples: 25
183
+ fov_pitch_samples: 20
184
+ fov_depth_samples: 20
185
+ pose_preselect_topk: 64
186
+ plucker_grid_h: 4
187
+ plucker_grid_w: 4
188
+ plucker_focal_length: 0.35
189
+ compress:
190
+ downsample_ratio: 2
191
+ stage_policy:
192
+ noise_bucket_logging: true
193
+ eval_ablation:
194
+ enabled: false
195
+ branch: A_plus_D_plus_R_normal
196
+ generated_history_proxy:
197
+ enabled: true
198
+ start_step: 40000
199
+ ramp_steps: 40000
200
+ max_prob: 0.25
201
+ noise_std: 0.25
202
+ dropout_prob: 0.0
203
+ injection:
204
+ memory_hidden_size: 512
205
+ cache:
206
+ enabled: true
207
+ device: cpu
208
+ keep_raw_latents: all
209
+ keep_compressed_records: true
210
+ keep_prefix_anchors: true
211
+ eviction_policy: none
212
+ no_evict: true
213
+ clear_between_videos: true
214
+ max_records: null
215
+ on_capacity_exceeded: warn
216
+ checkpoint:
217
+ strict_dememwm_eval_load: true
218
+ n_tokens: 4
219
+ debug: false
220
+ wandb:
221
+ entity: turlin
222
+ project: worldmem
223
+ mode: offline
224
+ resume: null
225
+ load: null
226
+ auto_resume: true
227
+ resume_ckpt_path: null
228
+ name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
229
+ output_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
230
+ customized_load: true
231
+ seperate_load: true
232
+ diffusion_model_path: /share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors
233
+ vae_path: /share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors
234
+ only_tune_memory: false
outputs/2026-06-01/15-41-01/.hydra/hydra.yaml ADDED
@@ -0,0 +1,232 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
4
+ sweep:
5
+ dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
6
+ subdir: ${hydra.job.num}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ formatters:
71
+ simple:
72
+ format: '[%(asctime)s][HYDRA] %(message)s'
73
+ handlers:
74
+ console:
75
+ class: logging.StreamHandler
76
+ formatter: simple
77
+ stream: ext://sys.stdout
78
+ root:
79
+ level: INFO
80
+ handlers:
81
+ - console
82
+ loggers:
83
+ logging_example:
84
+ level: DEBUG
85
+ disable_existing_loggers: false
86
+ job_logging:
87
+ version: 1
88
+ formatters:
89
+ simple:
90
+ format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
91
+ handlers:
92
+ console:
93
+ class: logging.StreamHandler
94
+ formatter: simple
95
+ stream: ext://sys.stdout
96
+ file:
97
+ class: logging.FileHandler
98
+ formatter: simple
99
+ filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
100
+ root:
101
+ level: INFO
102
+ handlers:
103
+ - console
104
+ - file
105
+ disable_existing_loggers: false
106
+ env: {}
107
+ mode: RUN
108
+ searchpath: []
109
+ callbacks: {}
110
+ output_subdir: .hydra
111
+ overrides:
112
+ hydra:
113
+ - hydra.mode=RUN
114
+ task:
115
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
116
+ - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
117
+ - auto_resume=true
118
+ - experiment.tasks=[training]
119
+ - algorithm=dememwm_memory_dit
120
+ - +customized_load=true
121
+ - +seperate_load=true
122
+ - +diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors
123
+ - +vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors
124
+ - +only_tune_memory=false
125
+ - dataset=video_minecraft_latent
126
+ - dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft
127
+ - dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features
128
+ - dataset.n_frames=1000
129
+ - +dataset.n_frames_valid=1100
130
+ - +dataset.customized_validation=true
131
+ - +dataset.wo_updown=false
132
+ - +dataset.angle_range=180
133
+ - +dataset.pos_range=8
134
+ - ++algorithm.n_tokens=4
135
+ - algorithm.x_shape=[16,18,32]
136
+ - ++algorithm.context_frames=100
137
+ - ++algorithm.log_video=true
138
+ - ++algorithm.diffusion.sampling_timesteps=20
139
+ - ++algorithm.dememwm.debug_force_all_streams=false
140
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
141
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
142
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
143
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
144
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
145
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
146
+ - ++algorithm.dememwm.anchor.enabled=true
147
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
148
+ - ++algorithm.dememwm.anchor.diverse_selection=true
149
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
150
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
151
+ - ++algorithm.dememwm.dynamic.enabled=true
152
+ - ++algorithm.dememwm.dynamic.recent_frames=4
153
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
154
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
155
+ - ++algorithm.dememwm.revisit.enabled=true
156
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
157
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
158
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
159
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
160
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
161
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
162
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
163
+ - ++algorithm.dememwm.revisit.max_frames=2
164
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
165
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
166
+ - ++algorithm.dememwm.cache.enabled=true
167
+ - ++algorithm.dememwm.cache.device=cpu
168
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
169
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
170
+ - ++algorithm.dememwm.cache.eviction_policy=none
171
+ - ++algorithm.dememwm.cache.no_evict=true
172
+ - ++algorithm.dememwm.cache.clear_between_videos=true
173
+ - ++algorithm.dememwm.cache.max_records=null
174
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
175
+ - ++algorithm.dememwm.curriculum.enabled=true
176
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
177
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
178
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
179
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
180
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
181
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
182
+ - experiment.training.batch_size=16
183
+ - experiment.training.optim.accumulate_grad_batches=1
184
+ - experiment.validation.batch_size=1
185
+ - experiment.validation.limit_batch=2
186
+ - experiment.training.checkpointing.every_n_train_steps=2000
187
+ - experiment.validation.val_every_n_step=2000
188
+ - experiment.training.max_steps=350000
189
+ job:
190
+ name: main
191
+ chdir: null
192
+ 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
193
+ id: ???
194
+ num: ???
195
+ config_name: training
196
+ env_set: {}
197
+ env_copy: []
198
+ config:
199
+ override_dirname:
200
+ kv_sep: '='
201
+ item_sep: ','
202
+ exclude_keys: []
203
+ runtime:
204
+ version: 1.3.2
205
+ version_base: '1.3'
206
+ cwd: /proj/cvl/users/x_fahkh2/DeMemWM
207
+ config_sources:
208
+ - path: hydra.conf
209
+ schema: pkg
210
+ provider: hydra
211
+ - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations
212
+ schema: file
213
+ provider: main
214
+ - path: ''
215
+ schema: structured
216
+ provider: schema
217
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-01/15-41-01
218
+ choices:
219
+ cluster: null
220
+ algorithm: dememwm_memory_dit
221
+ dataset: video_minecraft_latent
222
+ experiment: exp_video
223
+ hydra/env: default
224
+ hydra/callbacks: null
225
+ hydra/job_logging: default
226
+ hydra/hydra_logging: default
227
+ hydra/hydra_help: default
228
+ hydra/help: default
229
+ hydra/sweeper: basic
230
+ hydra/launcher: basic
231
+ hydra/output: default
232
+ verbose: false
outputs/2026-06-01/15-41-01/.hydra/overrides.yaml ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
2
+ - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
3
+ - auto_resume=true
4
+ - experiment.tasks=[training]
5
+ - algorithm=dememwm_memory_dit
6
+ - +customized_load=true
7
+ - +seperate_load=true
8
+ - +diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors
9
+ - +vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors
10
+ - +only_tune_memory=false
11
+ - dataset=video_minecraft_latent
12
+ - dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft
13
+ - dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features
14
+ - dataset.n_frames=1000
15
+ - +dataset.n_frames_valid=1100
16
+ - +dataset.customized_validation=true
17
+ - +dataset.wo_updown=false
18
+ - +dataset.angle_range=180
19
+ - +dataset.pos_range=8
20
+ - ++algorithm.n_tokens=4
21
+ - algorithm.x_shape=[16,18,32]
22
+ - ++algorithm.context_frames=100
23
+ - ++algorithm.log_video=true
24
+ - ++algorithm.diffusion.sampling_timesteps=20
25
+ - ++algorithm.dememwm.debug_force_all_streams=false
26
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
27
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
28
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
29
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
30
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
31
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
32
+ - ++algorithm.dememwm.anchor.enabled=true
33
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
34
+ - ++algorithm.dememwm.anchor.diverse_selection=true
35
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
36
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
37
+ - ++algorithm.dememwm.dynamic.enabled=true
38
+ - ++algorithm.dememwm.dynamic.recent_frames=4
39
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
40
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
41
+ - ++algorithm.dememwm.revisit.enabled=true
42
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
43
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
44
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
45
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
46
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
47
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
48
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
49
+ - ++algorithm.dememwm.revisit.max_frames=2
50
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
51
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
52
+ - ++algorithm.dememwm.cache.enabled=true
53
+ - ++algorithm.dememwm.cache.device=cpu
54
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
55
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
56
+ - ++algorithm.dememwm.cache.eviction_policy=none
57
+ - ++algorithm.dememwm.cache.no_evict=true
58
+ - ++algorithm.dememwm.cache.clear_between_videos=true
59
+ - ++algorithm.dememwm.cache.max_records=null
60
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
61
+ - ++algorithm.dememwm.curriculum.enabled=true
62
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
63
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
64
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
65
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
66
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
67
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
68
+ - experiment.training.batch_size=16
69
+ - experiment.training.optim.accumulate_grad_batches=1
70
+ - experiment.validation.batch_size=1
71
+ - experiment.validation.limit_batch=2
72
+ - experiment.training.checkpointing.every_n_train_steps=2000
73
+ - experiment.validation.val_every_n_step=2000
74
+ - experiment.training.max_steps=350000
outputs/2026-06-01/15-41-01/main.log ADDED
File without changes
outputs/2026-06-01/15-41-06/.hydra/config.yaml ADDED
@@ -0,0 +1,234 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment:
2
+ debug: ${debug}
3
+ tasks:
4
+ - training
5
+ num_nodes: 1
6
+ training:
7
+ precision: 16-mixed
8
+ compile: false
9
+ lr: 2.0e-05
10
+ batch_size: 16
11
+ max_epochs: -1
12
+ max_steps: 350000
13
+ max_time: null
14
+ data:
15
+ num_workers: 4
16
+ shuffle: true
17
+ optim:
18
+ accumulate_grad_batches: 1
19
+ gradient_clip_val: 1.0
20
+ checkpointing:
21
+ every_n_train_steps: 2000
22
+ every_n_epochs: null
23
+ train_time_interval: null
24
+ save_last: true
25
+ enable_version_counter: false
26
+ validation:
27
+ precision: 16-mixed
28
+ compile: false
29
+ batch_size: 1
30
+ val_every_n_step: 2000
31
+ val_every_n_epoch: null
32
+ limit_batch: 2
33
+ inference_mode: true
34
+ data:
35
+ num_workers: 4
36
+ shuffle: false
37
+ test:
38
+ precision: 16-mixed
39
+ compile: false
40
+ batch_size: 1
41
+ limit_batch: 1
42
+ data:
43
+ num_workers: 4
44
+ shuffle: false
45
+ logging:
46
+ metrics: null
47
+ dataset:
48
+ debug: ${debug}
49
+ metadata: data/${dataset.name}/metadata.json
50
+ data_mean: 0.5
51
+ data_std: 0.5
52
+ save_dir: /share_1/users/bonan_ding/worldmem_data/minecraft
53
+ n_frames: 1000
54
+ context_length: 1
55
+ resolution: 128
56
+ observation_shape:
57
+ - 3
58
+ - ${dataset.resolution}
59
+ - ${dataset.resolution}
60
+ external_cond_dim: 0
61
+ validation_multiplier: 1
62
+ frame_skip: 1
63
+ action_cond_dim: 25
64
+ _name: video_minecraft_latent
65
+ precomputed_feature_dir: /share_1/users/bonan_ding/worldmem_data/minecraft/vae_features
66
+ n_frames_valid: 1100
67
+ customized_validation: true
68
+ wo_updown: false
69
+ angle_range: 180
70
+ pos_range: 8
71
+ algorithm:
72
+ debug: ${debug}
73
+ lr: ${experiment.training.lr}
74
+ x_shape:
75
+ - 16
76
+ - 18
77
+ - 32
78
+ frame_stack: 1
79
+ frame_skip: ${dataset.frame_skip}
80
+ data_mean: ${dataset.data_mean}
81
+ data_std: ${dataset.data_std}
82
+ external_cond_dim: 0
83
+ context_frames: 100
84
+ weight_decay: 0.002
85
+ warmup_steps: 1000
86
+ optimizer_beta:
87
+ - 0.9
88
+ - 0.99
89
+ uncertainty_scale: 1
90
+ guidance_scale: 0.0
91
+ chunk_size: 1
92
+ scheduling_matrix: autoregressive
93
+ noise_level: random_all
94
+ causal: true
95
+ diffusion:
96
+ objective: pred_v
97
+ beta_schedule: sigmoid
98
+ schedule_fn_kwargs: {}
99
+ clip_noise: 20.0
100
+ use_snr: false
101
+ use_cum_snr: false
102
+ use_fused_snr: true
103
+ snr_clip: 5.0
104
+ cum_snr_decay: 0.96
105
+ timesteps: 1000
106
+ sampling_timesteps: 20
107
+ ddim_sampling_eta: 0.0
108
+ stabilization_level: 15
109
+ architecture:
110
+ network_size: 64
111
+ attn_heads: 4
112
+ attn_dim_head: 64
113
+ dim_mults:
114
+ - 1
115
+ - 2
116
+ - 4
117
+ - 8
118
+ resolution: ${dataset.resolution}
119
+ attn_resolutions:
120
+ - 16
121
+ - 32
122
+ - 64
123
+ - 128
124
+ use_init_temporal_attn: true
125
+ use_linear_attn: true
126
+ time_emb_type: rotary
127
+ n_frames: ${dataset.n_frames}
128
+ metadata: ${dataset.metadata}
129
+ action_cond_dim: 25
130
+ use_plucker: true
131
+ _name: dememwm_memory_dit
132
+ memory_token_cross_attention: true
133
+ memory_cross_attn_layers:
134
+ - 3
135
+ - 7
136
+ - 11
137
+ - 15
138
+ pose_cond_dim: 5
139
+ log_video: true
140
+ dememwm:
141
+ enabled: true
142
+ training_stage: stage_1
143
+ debug_force_all_streams: false
144
+ curriculum:
145
+ enabled: true
146
+ full_stage_start_step: 20000
147
+ freeze_vae: true
148
+ dit_freeze:
149
+ enabled: true
150
+ lr:
151
+ dememwm_modules: 4.0e-05
152
+ memory_adapters: 4.0e-05
153
+ full_dit: 1.0e-05
154
+ spatial_kernel_size: 3
155
+ anchor:
156
+ enabled: true
157
+ anchor_indices:
158
+ - 0
159
+ - 1
160
+ - 2
161
+ - 3
162
+ allow_generated_as_anchor: false
163
+ diverse_selection: true
164
+ compress:
165
+ downsample_ratio: 2
166
+ dynamic:
167
+ enabled: true
168
+ recent_frames: 4
169
+ conv_kernel_t: 3
170
+ conv_stride_t: 2
171
+ compress:
172
+ downsample_ratio: 2
173
+ revisit:
174
+ enabled: true
175
+ deterministic_pose_retrieval: true
176
+ fov_overlap_threshold: 0.6
177
+ plucker_weight: 0.1
178
+ max_frames: 2
179
+ fov_half_h: 52.5
180
+ fov_half_v: 37.5
181
+ fov_radius: 30.0
182
+ fov_yaw_samples: 25
183
+ fov_pitch_samples: 20
184
+ fov_depth_samples: 20
185
+ pose_preselect_topk: 64
186
+ plucker_grid_h: 4
187
+ plucker_grid_w: 4
188
+ plucker_focal_length: 0.35
189
+ compress:
190
+ downsample_ratio: 2
191
+ stage_policy:
192
+ noise_bucket_logging: true
193
+ eval_ablation:
194
+ enabled: false
195
+ branch: A_plus_D_plus_R_normal
196
+ generated_history_proxy:
197
+ enabled: true
198
+ start_step: 40000
199
+ ramp_steps: 40000
200
+ max_prob: 0.25
201
+ noise_std: 0.25
202
+ dropout_prob: 0.0
203
+ injection:
204
+ memory_hidden_size: 512
205
+ cache:
206
+ enabled: true
207
+ device: cpu
208
+ keep_raw_latents: all
209
+ keep_compressed_records: true
210
+ keep_prefix_anchors: true
211
+ eviction_policy: none
212
+ no_evict: true
213
+ clear_between_videos: true
214
+ max_records: null
215
+ on_capacity_exceeded: warn
216
+ checkpoint:
217
+ strict_dememwm_eval_load: true
218
+ n_tokens: 4
219
+ debug: false
220
+ wandb:
221
+ entity: turlin
222
+ project: worldmem
223
+ mode: offline
224
+ resume: null
225
+ load: null
226
+ auto_resume: true
227
+ resume_ckpt_path: null
228
+ name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
229
+ output_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
230
+ customized_load: true
231
+ seperate_load: true
232
+ diffusion_model_path: /share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors
233
+ vae_path: /share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors
234
+ only_tune_memory: false
outputs/2026-06-01/15-41-06/.hydra/hydra.yaml ADDED
@@ -0,0 +1,232 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
4
+ sweep:
5
+ dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
6
+ subdir: ${hydra.job.num}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ formatters:
71
+ simple:
72
+ format: '[%(asctime)s][HYDRA] %(message)s'
73
+ handlers:
74
+ console:
75
+ class: logging.StreamHandler
76
+ formatter: simple
77
+ stream: ext://sys.stdout
78
+ root:
79
+ level: INFO
80
+ handlers:
81
+ - console
82
+ loggers:
83
+ logging_example:
84
+ level: DEBUG
85
+ disable_existing_loggers: false
86
+ job_logging:
87
+ version: 1
88
+ formatters:
89
+ simple:
90
+ format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
91
+ handlers:
92
+ console:
93
+ class: logging.StreamHandler
94
+ formatter: simple
95
+ stream: ext://sys.stdout
96
+ file:
97
+ class: logging.FileHandler
98
+ formatter: simple
99
+ filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
100
+ root:
101
+ level: INFO
102
+ handlers:
103
+ - console
104
+ - file
105
+ disable_existing_loggers: false
106
+ env: {}
107
+ mode: RUN
108
+ searchpath: []
109
+ callbacks: {}
110
+ output_subdir: .hydra
111
+ overrides:
112
+ hydra:
113
+ - hydra.mode=RUN
114
+ task:
115
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
116
+ - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
117
+ - auto_resume=true
118
+ - experiment.tasks=[training]
119
+ - algorithm=dememwm_memory_dit
120
+ - +customized_load=true
121
+ - +seperate_load=true
122
+ - +diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors
123
+ - +vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors
124
+ - +only_tune_memory=false
125
+ - dataset=video_minecraft_latent
126
+ - dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft
127
+ - dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features
128
+ - dataset.n_frames=1000
129
+ - +dataset.n_frames_valid=1100
130
+ - +dataset.customized_validation=true
131
+ - +dataset.wo_updown=false
132
+ - +dataset.angle_range=180
133
+ - +dataset.pos_range=8
134
+ - ++algorithm.n_tokens=4
135
+ - algorithm.x_shape=[16,18,32]
136
+ - ++algorithm.context_frames=100
137
+ - ++algorithm.log_video=true
138
+ - ++algorithm.diffusion.sampling_timesteps=20
139
+ - ++algorithm.dememwm.debug_force_all_streams=false
140
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
141
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
142
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
143
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
144
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
145
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
146
+ - ++algorithm.dememwm.anchor.enabled=true
147
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
148
+ - ++algorithm.dememwm.anchor.diverse_selection=true
149
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
150
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
151
+ - ++algorithm.dememwm.dynamic.enabled=true
152
+ - ++algorithm.dememwm.dynamic.recent_frames=4
153
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
154
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
155
+ - ++algorithm.dememwm.revisit.enabled=true
156
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
157
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
158
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
159
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
160
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
161
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
162
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
163
+ - ++algorithm.dememwm.revisit.max_frames=2
164
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
165
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
166
+ - ++algorithm.dememwm.cache.enabled=true
167
+ - ++algorithm.dememwm.cache.device=cpu
168
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
169
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
170
+ - ++algorithm.dememwm.cache.eviction_policy=none
171
+ - ++algorithm.dememwm.cache.no_evict=true
172
+ - ++algorithm.dememwm.cache.clear_between_videos=true
173
+ - ++algorithm.dememwm.cache.max_records=null
174
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
175
+ - ++algorithm.dememwm.curriculum.enabled=true
176
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
177
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
178
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
179
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
180
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
181
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
182
+ - experiment.training.batch_size=16
183
+ - experiment.training.optim.accumulate_grad_batches=1
184
+ - experiment.validation.batch_size=1
185
+ - experiment.validation.limit_batch=2
186
+ - experiment.training.checkpointing.every_n_train_steps=2000
187
+ - experiment.validation.val_every_n_step=2000
188
+ - experiment.training.max_steps=350000
189
+ job:
190
+ name: main
191
+ chdir: null
192
+ 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
193
+ id: ???
194
+ num: ???
195
+ config_name: training
196
+ env_set: {}
197
+ env_copy: []
198
+ config:
199
+ override_dirname:
200
+ kv_sep: '='
201
+ item_sep: ','
202
+ exclude_keys: []
203
+ runtime:
204
+ version: 1.3.2
205
+ version_base: '1.3'
206
+ cwd: /proj/cvl/users/x_fahkh2/DeMemWM
207
+ config_sources:
208
+ - path: hydra.conf
209
+ schema: pkg
210
+ provider: hydra
211
+ - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations
212
+ schema: file
213
+ provider: main
214
+ - path: ''
215
+ schema: structured
216
+ provider: schema
217
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-01/15-41-06
218
+ choices:
219
+ cluster: null
220
+ algorithm: dememwm_memory_dit
221
+ dataset: video_minecraft_latent
222
+ experiment: exp_video
223
+ hydra/env: default
224
+ hydra/callbacks: null
225
+ hydra/job_logging: default
226
+ hydra/hydra_logging: default
227
+ hydra/hydra_help: default
228
+ hydra/help: default
229
+ hydra/sweeper: basic
230
+ hydra/launcher: basic
231
+ hydra/output: default
232
+ verbose: false
outputs/2026-06-01/15-41-06/.hydra/overrides.yaml ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
2
+ - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
3
+ - auto_resume=true
4
+ - experiment.tasks=[training]
5
+ - algorithm=dememwm_memory_dit
6
+ - +customized_load=true
7
+ - +seperate_load=true
8
+ - +diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors
9
+ - +vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors
10
+ - +only_tune_memory=false
11
+ - dataset=video_minecraft_latent
12
+ - dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft
13
+ - dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features
14
+ - dataset.n_frames=1000
15
+ - +dataset.n_frames_valid=1100
16
+ - +dataset.customized_validation=true
17
+ - +dataset.wo_updown=false
18
+ - +dataset.angle_range=180
19
+ - +dataset.pos_range=8
20
+ - ++algorithm.n_tokens=4
21
+ - algorithm.x_shape=[16,18,32]
22
+ - ++algorithm.context_frames=100
23
+ - ++algorithm.log_video=true
24
+ - ++algorithm.diffusion.sampling_timesteps=20
25
+ - ++algorithm.dememwm.debug_force_all_streams=false
26
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
27
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
28
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
29
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
30
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
31
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
32
+ - ++algorithm.dememwm.anchor.enabled=true
33
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
34
+ - ++algorithm.dememwm.anchor.diverse_selection=true
35
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
36
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
37
+ - ++algorithm.dememwm.dynamic.enabled=true
38
+ - ++algorithm.dememwm.dynamic.recent_frames=4
39
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
40
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
41
+ - ++algorithm.dememwm.revisit.enabled=true
42
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
43
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
44
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
45
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
46
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
47
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
48
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
49
+ - ++algorithm.dememwm.revisit.max_frames=2
50
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
51
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
52
+ - ++algorithm.dememwm.cache.enabled=true
53
+ - ++algorithm.dememwm.cache.device=cpu
54
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
55
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
56
+ - ++algorithm.dememwm.cache.eviction_policy=none
57
+ - ++algorithm.dememwm.cache.no_evict=true
58
+ - ++algorithm.dememwm.cache.clear_between_videos=true
59
+ - ++algorithm.dememwm.cache.max_records=null
60
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
61
+ - ++algorithm.dememwm.curriculum.enabled=true
62
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
63
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
64
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
65
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
66
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
67
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
68
+ - experiment.training.batch_size=16
69
+ - experiment.training.optim.accumulate_grad_batches=1
70
+ - experiment.validation.batch_size=1
71
+ - experiment.validation.limit_batch=2
72
+ - experiment.training.checkpointing.every_n_train_steps=2000
73
+ - experiment.validation.val_every_n_step=2000
74
+ - experiment.training.max_steps=350000
outputs/2026-06-01/15-41-06/main.log ADDED
File without changes
outputs/2026-06-01/15-41-07/.hydra/config.yaml ADDED
@@ -0,0 +1,234 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment:
2
+ debug: ${debug}
3
+ tasks:
4
+ - training
5
+ num_nodes: 1
6
+ training:
7
+ precision: 16-mixed
8
+ compile: false
9
+ lr: 2.0e-05
10
+ batch_size: 16
11
+ max_epochs: -1
12
+ max_steps: 350000
13
+ max_time: null
14
+ data:
15
+ num_workers: 4
16
+ shuffle: true
17
+ optim:
18
+ accumulate_grad_batches: 1
19
+ gradient_clip_val: 1.0
20
+ checkpointing:
21
+ every_n_train_steps: 2000
22
+ every_n_epochs: null
23
+ train_time_interval: null
24
+ save_last: true
25
+ enable_version_counter: false
26
+ validation:
27
+ precision: 16-mixed
28
+ compile: false
29
+ batch_size: 1
30
+ val_every_n_step: 2000
31
+ val_every_n_epoch: null
32
+ limit_batch: 2
33
+ inference_mode: true
34
+ data:
35
+ num_workers: 4
36
+ shuffle: false
37
+ test:
38
+ precision: 16-mixed
39
+ compile: false
40
+ batch_size: 1
41
+ limit_batch: 1
42
+ data:
43
+ num_workers: 4
44
+ shuffle: false
45
+ logging:
46
+ metrics: null
47
+ dataset:
48
+ debug: ${debug}
49
+ metadata: data/${dataset.name}/metadata.json
50
+ data_mean: 0.5
51
+ data_std: 0.5
52
+ save_dir: /share_1/users/bonan_ding/worldmem_data/minecraft
53
+ n_frames: 1000
54
+ context_length: 1
55
+ resolution: 128
56
+ observation_shape:
57
+ - 3
58
+ - ${dataset.resolution}
59
+ - ${dataset.resolution}
60
+ external_cond_dim: 0
61
+ validation_multiplier: 1
62
+ frame_skip: 1
63
+ action_cond_dim: 25
64
+ _name: video_minecraft_latent
65
+ precomputed_feature_dir: /share_1/users/bonan_ding/worldmem_data/minecraft/vae_features
66
+ n_frames_valid: 1100
67
+ customized_validation: true
68
+ wo_updown: false
69
+ angle_range: 180
70
+ pos_range: 8
71
+ algorithm:
72
+ debug: ${debug}
73
+ lr: ${experiment.training.lr}
74
+ x_shape:
75
+ - 16
76
+ - 18
77
+ - 32
78
+ frame_stack: 1
79
+ frame_skip: ${dataset.frame_skip}
80
+ data_mean: ${dataset.data_mean}
81
+ data_std: ${dataset.data_std}
82
+ external_cond_dim: 0
83
+ context_frames: 100
84
+ weight_decay: 0.002
85
+ warmup_steps: 1000
86
+ optimizer_beta:
87
+ - 0.9
88
+ - 0.99
89
+ uncertainty_scale: 1
90
+ guidance_scale: 0.0
91
+ chunk_size: 1
92
+ scheduling_matrix: autoregressive
93
+ noise_level: random_all
94
+ causal: true
95
+ diffusion:
96
+ objective: pred_v
97
+ beta_schedule: sigmoid
98
+ schedule_fn_kwargs: {}
99
+ clip_noise: 20.0
100
+ use_snr: false
101
+ use_cum_snr: false
102
+ use_fused_snr: true
103
+ snr_clip: 5.0
104
+ cum_snr_decay: 0.96
105
+ timesteps: 1000
106
+ sampling_timesteps: 20
107
+ ddim_sampling_eta: 0.0
108
+ stabilization_level: 15
109
+ architecture:
110
+ network_size: 64
111
+ attn_heads: 4
112
+ attn_dim_head: 64
113
+ dim_mults:
114
+ - 1
115
+ - 2
116
+ - 4
117
+ - 8
118
+ resolution: ${dataset.resolution}
119
+ attn_resolutions:
120
+ - 16
121
+ - 32
122
+ - 64
123
+ - 128
124
+ use_init_temporal_attn: true
125
+ use_linear_attn: true
126
+ time_emb_type: rotary
127
+ n_frames: ${dataset.n_frames}
128
+ metadata: ${dataset.metadata}
129
+ action_cond_dim: 25
130
+ use_plucker: true
131
+ _name: dememwm_memory_dit
132
+ memory_token_cross_attention: true
133
+ memory_cross_attn_layers:
134
+ - 3
135
+ - 7
136
+ - 11
137
+ - 15
138
+ pose_cond_dim: 5
139
+ log_video: true
140
+ dememwm:
141
+ enabled: true
142
+ training_stage: stage_1
143
+ debug_force_all_streams: false
144
+ curriculum:
145
+ enabled: true
146
+ full_stage_start_step: 20000
147
+ freeze_vae: true
148
+ dit_freeze:
149
+ enabled: true
150
+ lr:
151
+ dememwm_modules: 4.0e-05
152
+ memory_adapters: 4.0e-05
153
+ full_dit: 1.0e-05
154
+ spatial_kernel_size: 3
155
+ anchor:
156
+ enabled: true
157
+ anchor_indices:
158
+ - 0
159
+ - 1
160
+ - 2
161
+ - 3
162
+ allow_generated_as_anchor: false
163
+ diverse_selection: true
164
+ compress:
165
+ downsample_ratio: 2
166
+ dynamic:
167
+ enabled: true
168
+ recent_frames: 4
169
+ conv_kernel_t: 3
170
+ conv_stride_t: 2
171
+ compress:
172
+ downsample_ratio: 2
173
+ revisit:
174
+ enabled: true
175
+ deterministic_pose_retrieval: true
176
+ fov_overlap_threshold: 0.6
177
+ plucker_weight: 0.1
178
+ max_frames: 2
179
+ fov_half_h: 52.5
180
+ fov_half_v: 37.5
181
+ fov_radius: 30.0
182
+ fov_yaw_samples: 25
183
+ fov_pitch_samples: 20
184
+ fov_depth_samples: 20
185
+ pose_preselect_topk: 64
186
+ plucker_grid_h: 4
187
+ plucker_grid_w: 4
188
+ plucker_focal_length: 0.35
189
+ compress:
190
+ downsample_ratio: 2
191
+ stage_policy:
192
+ noise_bucket_logging: true
193
+ eval_ablation:
194
+ enabled: false
195
+ branch: A_plus_D_plus_R_normal
196
+ generated_history_proxy:
197
+ enabled: true
198
+ start_step: 40000
199
+ ramp_steps: 40000
200
+ max_prob: 0.25
201
+ noise_std: 0.25
202
+ dropout_prob: 0.0
203
+ injection:
204
+ memory_hidden_size: 512
205
+ cache:
206
+ enabled: true
207
+ device: cpu
208
+ keep_raw_latents: all
209
+ keep_compressed_records: true
210
+ keep_prefix_anchors: true
211
+ eviction_policy: none
212
+ no_evict: true
213
+ clear_between_videos: true
214
+ max_records: null
215
+ on_capacity_exceeded: warn
216
+ checkpoint:
217
+ strict_dememwm_eval_load: true
218
+ n_tokens: 4
219
+ debug: false
220
+ wandb:
221
+ entity: turlin
222
+ project: worldmem
223
+ mode: offline
224
+ resume: null
225
+ load: null
226
+ auto_resume: true
227
+ resume_ckpt_path: null
228
+ name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
229
+ output_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
230
+ customized_load: true
231
+ seperate_load: true
232
+ diffusion_model_path: /share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors
233
+ vae_path: /share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors
234
+ only_tune_memory: false
outputs/2026-06-01/15-41-07/.hydra/hydra.yaml ADDED
@@ -0,0 +1,232 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
4
+ sweep:
5
+ dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
6
+ subdir: ${hydra.job.num}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ formatters:
71
+ simple:
72
+ format: '[%(asctime)s][HYDRA] %(message)s'
73
+ handlers:
74
+ console:
75
+ class: logging.StreamHandler
76
+ formatter: simple
77
+ stream: ext://sys.stdout
78
+ root:
79
+ level: INFO
80
+ handlers:
81
+ - console
82
+ loggers:
83
+ logging_example:
84
+ level: DEBUG
85
+ disable_existing_loggers: false
86
+ job_logging:
87
+ version: 1
88
+ formatters:
89
+ simple:
90
+ format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
91
+ handlers:
92
+ console:
93
+ class: logging.StreamHandler
94
+ formatter: simple
95
+ stream: ext://sys.stdout
96
+ file:
97
+ class: logging.FileHandler
98
+ formatter: simple
99
+ filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
100
+ root:
101
+ level: INFO
102
+ handlers:
103
+ - console
104
+ - file
105
+ disable_existing_loggers: false
106
+ env: {}
107
+ mode: RUN
108
+ searchpath: []
109
+ callbacks: {}
110
+ output_subdir: .hydra
111
+ overrides:
112
+ hydra:
113
+ - hydra.mode=RUN
114
+ task:
115
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
116
+ - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
117
+ - auto_resume=true
118
+ - experiment.tasks=[training]
119
+ - algorithm=dememwm_memory_dit
120
+ - +customized_load=true
121
+ - +seperate_load=true
122
+ - +diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors
123
+ - +vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors
124
+ - +only_tune_memory=false
125
+ - dataset=video_minecraft_latent
126
+ - dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft
127
+ - dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features
128
+ - dataset.n_frames=1000
129
+ - +dataset.n_frames_valid=1100
130
+ - +dataset.customized_validation=true
131
+ - +dataset.wo_updown=false
132
+ - +dataset.angle_range=180
133
+ - +dataset.pos_range=8
134
+ - ++algorithm.n_tokens=4
135
+ - algorithm.x_shape=[16,18,32]
136
+ - ++algorithm.context_frames=100
137
+ - ++algorithm.log_video=true
138
+ - ++algorithm.diffusion.sampling_timesteps=20
139
+ - ++algorithm.dememwm.debug_force_all_streams=false
140
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
141
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
142
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
143
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
144
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
145
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
146
+ - ++algorithm.dememwm.anchor.enabled=true
147
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
148
+ - ++algorithm.dememwm.anchor.diverse_selection=true
149
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
150
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
151
+ - ++algorithm.dememwm.dynamic.enabled=true
152
+ - ++algorithm.dememwm.dynamic.recent_frames=4
153
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
154
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
155
+ - ++algorithm.dememwm.revisit.enabled=true
156
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
157
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
158
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
159
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
160
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
161
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
162
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
163
+ - ++algorithm.dememwm.revisit.max_frames=2
164
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
165
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
166
+ - ++algorithm.dememwm.cache.enabled=true
167
+ - ++algorithm.dememwm.cache.device=cpu
168
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
169
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
170
+ - ++algorithm.dememwm.cache.eviction_policy=none
171
+ - ++algorithm.dememwm.cache.no_evict=true
172
+ - ++algorithm.dememwm.cache.clear_between_videos=true
173
+ - ++algorithm.dememwm.cache.max_records=null
174
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
175
+ - ++algorithm.dememwm.curriculum.enabled=true
176
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
177
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
178
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
179
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
180
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
181
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
182
+ - experiment.training.batch_size=16
183
+ - experiment.training.optim.accumulate_grad_batches=1
184
+ - experiment.validation.batch_size=1
185
+ - experiment.validation.limit_batch=2
186
+ - experiment.training.checkpointing.every_n_train_steps=2000
187
+ - experiment.validation.val_every_n_step=2000
188
+ - experiment.training.max_steps=350000
189
+ job:
190
+ name: main
191
+ chdir: null
192
+ 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
193
+ id: ???
194
+ num: ???
195
+ config_name: training
196
+ env_set: {}
197
+ env_copy: []
198
+ config:
199
+ override_dirname:
200
+ kv_sep: '='
201
+ item_sep: ','
202
+ exclude_keys: []
203
+ runtime:
204
+ version: 1.3.2
205
+ version_base: '1.3'
206
+ cwd: /proj/cvl/users/x_fahkh2/DeMemWM
207
+ config_sources:
208
+ - path: hydra.conf
209
+ schema: pkg
210
+ provider: hydra
211
+ - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations
212
+ schema: file
213
+ provider: main
214
+ - path: ''
215
+ schema: structured
216
+ provider: schema
217
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-01/15-41-07
218
+ choices:
219
+ cluster: null
220
+ algorithm: dememwm_memory_dit
221
+ dataset: video_minecraft_latent
222
+ experiment: exp_video
223
+ hydra/env: default
224
+ hydra/callbacks: null
225
+ hydra/job_logging: default
226
+ hydra/hydra_logging: default
227
+ hydra/hydra_help: default
228
+ hydra/help: default
229
+ hydra/sweeper: basic
230
+ hydra/launcher: basic
231
+ hydra/output: default
232
+ verbose: false
outputs/2026-06-01/15-41-07/.hydra/overrides.yaml ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
2
+ - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
3
+ - auto_resume=true
4
+ - experiment.tasks=[training]
5
+ - algorithm=dememwm_memory_dit
6
+ - +customized_load=true
7
+ - +seperate_load=true
8
+ - +diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors
9
+ - +vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors
10
+ - +only_tune_memory=false
11
+ - dataset=video_minecraft_latent
12
+ - dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft
13
+ - dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features
14
+ - dataset.n_frames=1000
15
+ - +dataset.n_frames_valid=1100
16
+ - +dataset.customized_validation=true
17
+ - +dataset.wo_updown=false
18
+ - +dataset.angle_range=180
19
+ - +dataset.pos_range=8
20
+ - ++algorithm.n_tokens=4
21
+ - algorithm.x_shape=[16,18,32]
22
+ - ++algorithm.context_frames=100
23
+ - ++algorithm.log_video=true
24
+ - ++algorithm.diffusion.sampling_timesteps=20
25
+ - ++algorithm.dememwm.debug_force_all_streams=false
26
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
27
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
28
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
29
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
30
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
31
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
32
+ - ++algorithm.dememwm.anchor.enabled=true
33
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
34
+ - ++algorithm.dememwm.anchor.diverse_selection=true
35
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
36
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
37
+ - ++algorithm.dememwm.dynamic.enabled=true
38
+ - ++algorithm.dememwm.dynamic.recent_frames=4
39
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
40
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
41
+ - ++algorithm.dememwm.revisit.enabled=true
42
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
43
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
44
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
45
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
46
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
47
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
48
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
49
+ - ++algorithm.dememwm.revisit.max_frames=2
50
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
51
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
52
+ - ++algorithm.dememwm.cache.enabled=true
53
+ - ++algorithm.dememwm.cache.device=cpu
54
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
55
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
56
+ - ++algorithm.dememwm.cache.eviction_policy=none
57
+ - ++algorithm.dememwm.cache.no_evict=true
58
+ - ++algorithm.dememwm.cache.clear_between_videos=true
59
+ - ++algorithm.dememwm.cache.max_records=null
60
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
61
+ - ++algorithm.dememwm.curriculum.enabled=true
62
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
63
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
64
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
65
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
66
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
67
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
68
+ - experiment.training.batch_size=16
69
+ - experiment.training.optim.accumulate_grad_batches=1
70
+ - experiment.validation.batch_size=1
71
+ - experiment.validation.limit_batch=2
72
+ - experiment.training.checkpointing.every_n_train_steps=2000
73
+ - experiment.validation.val_every_n_step=2000
74
+ - experiment.training.max_steps=350000
outputs/2026-06-01/15-41-07/main.log ADDED
File without changes
outputs/2026-06-01/15-41-08/.hydra/config.yaml ADDED
@@ -0,0 +1,234 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment:
2
+ debug: ${debug}
3
+ tasks:
4
+ - training
5
+ num_nodes: 1
6
+ training:
7
+ precision: 16-mixed
8
+ compile: false
9
+ lr: 2.0e-05
10
+ batch_size: 16
11
+ max_epochs: -1
12
+ max_steps: 350000
13
+ max_time: null
14
+ data:
15
+ num_workers: 4
16
+ shuffle: true
17
+ optim:
18
+ accumulate_grad_batches: 1
19
+ gradient_clip_val: 1.0
20
+ checkpointing:
21
+ every_n_train_steps: 2000
22
+ every_n_epochs: null
23
+ train_time_interval: null
24
+ save_last: true
25
+ enable_version_counter: false
26
+ validation:
27
+ precision: 16-mixed
28
+ compile: false
29
+ batch_size: 1
30
+ val_every_n_step: 2000
31
+ val_every_n_epoch: null
32
+ limit_batch: 2
33
+ inference_mode: true
34
+ data:
35
+ num_workers: 4
36
+ shuffle: false
37
+ test:
38
+ precision: 16-mixed
39
+ compile: false
40
+ batch_size: 1
41
+ limit_batch: 1
42
+ data:
43
+ num_workers: 4
44
+ shuffle: false
45
+ logging:
46
+ metrics: null
47
+ dataset:
48
+ debug: ${debug}
49
+ metadata: data/${dataset.name}/metadata.json
50
+ data_mean: 0.5
51
+ data_std: 0.5
52
+ save_dir: /share_1/users/bonan_ding/worldmem_data/minecraft
53
+ n_frames: 1000
54
+ context_length: 1
55
+ resolution: 128
56
+ observation_shape:
57
+ - 3
58
+ - ${dataset.resolution}
59
+ - ${dataset.resolution}
60
+ external_cond_dim: 0
61
+ validation_multiplier: 1
62
+ frame_skip: 1
63
+ action_cond_dim: 25
64
+ _name: video_minecraft_latent
65
+ precomputed_feature_dir: /share_1/users/bonan_ding/worldmem_data/minecraft/vae_features
66
+ n_frames_valid: 1100
67
+ customized_validation: true
68
+ wo_updown: false
69
+ angle_range: 180
70
+ pos_range: 8
71
+ algorithm:
72
+ debug: ${debug}
73
+ lr: ${experiment.training.lr}
74
+ x_shape:
75
+ - 16
76
+ - 18
77
+ - 32
78
+ frame_stack: 1
79
+ frame_skip: ${dataset.frame_skip}
80
+ data_mean: ${dataset.data_mean}
81
+ data_std: ${dataset.data_std}
82
+ external_cond_dim: 0
83
+ context_frames: 100
84
+ weight_decay: 0.002
85
+ warmup_steps: 1000
86
+ optimizer_beta:
87
+ - 0.9
88
+ - 0.99
89
+ uncertainty_scale: 1
90
+ guidance_scale: 0.0
91
+ chunk_size: 1
92
+ scheduling_matrix: autoregressive
93
+ noise_level: random_all
94
+ causal: true
95
+ diffusion:
96
+ objective: pred_v
97
+ beta_schedule: sigmoid
98
+ schedule_fn_kwargs: {}
99
+ clip_noise: 20.0
100
+ use_snr: false
101
+ use_cum_snr: false
102
+ use_fused_snr: true
103
+ snr_clip: 5.0
104
+ cum_snr_decay: 0.96
105
+ timesteps: 1000
106
+ sampling_timesteps: 20
107
+ ddim_sampling_eta: 0.0
108
+ stabilization_level: 15
109
+ architecture:
110
+ network_size: 64
111
+ attn_heads: 4
112
+ attn_dim_head: 64
113
+ dim_mults:
114
+ - 1
115
+ - 2
116
+ - 4
117
+ - 8
118
+ resolution: ${dataset.resolution}
119
+ attn_resolutions:
120
+ - 16
121
+ - 32
122
+ - 64
123
+ - 128
124
+ use_init_temporal_attn: true
125
+ use_linear_attn: true
126
+ time_emb_type: rotary
127
+ n_frames: ${dataset.n_frames}
128
+ metadata: ${dataset.metadata}
129
+ action_cond_dim: 25
130
+ use_plucker: true
131
+ _name: dememwm_memory_dit
132
+ memory_token_cross_attention: true
133
+ memory_cross_attn_layers:
134
+ - 3
135
+ - 7
136
+ - 11
137
+ - 15
138
+ pose_cond_dim: 5
139
+ log_video: true
140
+ dememwm:
141
+ enabled: true
142
+ training_stage: stage_1
143
+ debug_force_all_streams: false
144
+ curriculum:
145
+ enabled: true
146
+ full_stage_start_step: 20000
147
+ freeze_vae: true
148
+ dit_freeze:
149
+ enabled: true
150
+ lr:
151
+ dememwm_modules: 4.0e-05
152
+ memory_adapters: 4.0e-05
153
+ full_dit: 1.0e-05
154
+ spatial_kernel_size: 3
155
+ anchor:
156
+ enabled: true
157
+ anchor_indices:
158
+ - 0
159
+ - 1
160
+ - 2
161
+ - 3
162
+ allow_generated_as_anchor: false
163
+ diverse_selection: true
164
+ compress:
165
+ downsample_ratio: 2
166
+ dynamic:
167
+ enabled: true
168
+ recent_frames: 4
169
+ conv_kernel_t: 3
170
+ conv_stride_t: 2
171
+ compress:
172
+ downsample_ratio: 2
173
+ revisit:
174
+ enabled: true
175
+ deterministic_pose_retrieval: true
176
+ fov_overlap_threshold: 0.6
177
+ plucker_weight: 0.1
178
+ max_frames: 2
179
+ fov_half_h: 52.5
180
+ fov_half_v: 37.5
181
+ fov_radius: 30.0
182
+ fov_yaw_samples: 25
183
+ fov_pitch_samples: 20
184
+ fov_depth_samples: 20
185
+ pose_preselect_topk: 64
186
+ plucker_grid_h: 4
187
+ plucker_grid_w: 4
188
+ plucker_focal_length: 0.35
189
+ compress:
190
+ downsample_ratio: 2
191
+ stage_policy:
192
+ noise_bucket_logging: true
193
+ eval_ablation:
194
+ enabled: false
195
+ branch: A_plus_D_plus_R_normal
196
+ generated_history_proxy:
197
+ enabled: true
198
+ start_step: 40000
199
+ ramp_steps: 40000
200
+ max_prob: 0.25
201
+ noise_std: 0.25
202
+ dropout_prob: 0.0
203
+ injection:
204
+ memory_hidden_size: 512
205
+ cache:
206
+ enabled: true
207
+ device: cpu
208
+ keep_raw_latents: all
209
+ keep_compressed_records: true
210
+ keep_prefix_anchors: true
211
+ eviction_policy: none
212
+ no_evict: true
213
+ clear_between_videos: true
214
+ max_records: null
215
+ on_capacity_exceeded: warn
216
+ checkpoint:
217
+ strict_dememwm_eval_load: true
218
+ n_tokens: 4
219
+ debug: false
220
+ wandb:
221
+ entity: turlin
222
+ project: worldmem
223
+ mode: offline
224
+ resume: null
225
+ load: null
226
+ auto_resume: true
227
+ resume_ckpt_path: null
228
+ name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
229
+ output_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
230
+ customized_load: true
231
+ seperate_load: true
232
+ diffusion_model_path: /share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors
233
+ vae_path: /share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors
234
+ only_tune_memory: false
outputs/2026-06-01/15-41-08/.hydra/hydra.yaml ADDED
@@ -0,0 +1,232 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
4
+ sweep:
5
+ dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
6
+ subdir: ${hydra.job.num}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ formatters:
71
+ simple:
72
+ format: '[%(asctime)s][HYDRA] %(message)s'
73
+ handlers:
74
+ console:
75
+ class: logging.StreamHandler
76
+ formatter: simple
77
+ stream: ext://sys.stdout
78
+ root:
79
+ level: INFO
80
+ handlers:
81
+ - console
82
+ loggers:
83
+ logging_example:
84
+ level: DEBUG
85
+ disable_existing_loggers: false
86
+ job_logging:
87
+ version: 1
88
+ formatters:
89
+ simple:
90
+ format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
91
+ handlers:
92
+ console:
93
+ class: logging.StreamHandler
94
+ formatter: simple
95
+ stream: ext://sys.stdout
96
+ file:
97
+ class: logging.FileHandler
98
+ formatter: simple
99
+ filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
100
+ root:
101
+ level: INFO
102
+ handlers:
103
+ - console
104
+ - file
105
+ disable_existing_loggers: false
106
+ env: {}
107
+ mode: RUN
108
+ searchpath: []
109
+ callbacks: {}
110
+ output_subdir: .hydra
111
+ overrides:
112
+ hydra:
113
+ - hydra.mode=RUN
114
+ task:
115
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
116
+ - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
117
+ - auto_resume=true
118
+ - experiment.tasks=[training]
119
+ - algorithm=dememwm_memory_dit
120
+ - +customized_load=true
121
+ - +seperate_load=true
122
+ - +diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors
123
+ - +vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors
124
+ - +only_tune_memory=false
125
+ - dataset=video_minecraft_latent
126
+ - dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft
127
+ - dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features
128
+ - dataset.n_frames=1000
129
+ - +dataset.n_frames_valid=1100
130
+ - +dataset.customized_validation=true
131
+ - +dataset.wo_updown=false
132
+ - +dataset.angle_range=180
133
+ - +dataset.pos_range=8
134
+ - ++algorithm.n_tokens=4
135
+ - algorithm.x_shape=[16,18,32]
136
+ - ++algorithm.context_frames=100
137
+ - ++algorithm.log_video=true
138
+ - ++algorithm.diffusion.sampling_timesteps=20
139
+ - ++algorithm.dememwm.debug_force_all_streams=false
140
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
141
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
142
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
143
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
144
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
145
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
146
+ - ++algorithm.dememwm.anchor.enabled=true
147
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
148
+ - ++algorithm.dememwm.anchor.diverse_selection=true
149
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
150
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
151
+ - ++algorithm.dememwm.dynamic.enabled=true
152
+ - ++algorithm.dememwm.dynamic.recent_frames=4
153
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
154
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
155
+ - ++algorithm.dememwm.revisit.enabled=true
156
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
157
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
158
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
159
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
160
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
161
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
162
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
163
+ - ++algorithm.dememwm.revisit.max_frames=2
164
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
165
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
166
+ - ++algorithm.dememwm.cache.enabled=true
167
+ - ++algorithm.dememwm.cache.device=cpu
168
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
169
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
170
+ - ++algorithm.dememwm.cache.eviction_policy=none
171
+ - ++algorithm.dememwm.cache.no_evict=true
172
+ - ++algorithm.dememwm.cache.clear_between_videos=true
173
+ - ++algorithm.dememwm.cache.max_records=null
174
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
175
+ - ++algorithm.dememwm.curriculum.enabled=true
176
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
177
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
178
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
179
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
180
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
181
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
182
+ - experiment.training.batch_size=16
183
+ - experiment.training.optim.accumulate_grad_batches=1
184
+ - experiment.validation.batch_size=1
185
+ - experiment.validation.limit_batch=2
186
+ - experiment.training.checkpointing.every_n_train_steps=2000
187
+ - experiment.validation.val_every_n_step=2000
188
+ - experiment.training.max_steps=350000
189
+ job:
190
+ name: main
191
+ chdir: null
192
+ 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
193
+ id: ???
194
+ num: ???
195
+ config_name: training
196
+ env_set: {}
197
+ env_copy: []
198
+ config:
199
+ override_dirname:
200
+ kv_sep: '='
201
+ item_sep: ','
202
+ exclude_keys: []
203
+ runtime:
204
+ version: 1.3.2
205
+ version_base: '1.3'
206
+ cwd: /proj/cvl/users/x_fahkh2/DeMemWM
207
+ config_sources:
208
+ - path: hydra.conf
209
+ schema: pkg
210
+ provider: hydra
211
+ - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations
212
+ schema: file
213
+ provider: main
214
+ - path: ''
215
+ schema: structured
216
+ provider: schema
217
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-01/15-41-08
218
+ choices:
219
+ cluster: null
220
+ algorithm: dememwm_memory_dit
221
+ dataset: video_minecraft_latent
222
+ experiment: exp_video
223
+ hydra/env: default
224
+ hydra/callbacks: null
225
+ hydra/job_logging: default
226
+ hydra/hydra_logging: default
227
+ hydra/hydra_help: default
228
+ hydra/help: default
229
+ hydra/sweeper: basic
230
+ hydra/launcher: basic
231
+ hydra/output: default
232
+ verbose: false
outputs/2026-06-01/15-41-08/.hydra/overrides.yaml ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
2
+ - +output_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
3
+ - auto_resume=true
4
+ - experiment.tasks=[training]
5
+ - algorithm=dememwm_memory_dit
6
+ - +customized_load=true
7
+ - +seperate_load=true
8
+ - +diffusion_model_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/oasis500m.safetensors
9
+ - +vae_path=/share_1/users/bonan_ding/WorldMem/open-oasis/checkpoints/vit-l-20.safetensors
10
+ - +only_tune_memory=false
11
+ - dataset=video_minecraft_latent
12
+ - dataset.save_dir=/share_1/users/bonan_ding/worldmem_data/minecraft
13
+ - dataset.precomputed_feature_dir=/share_1/users/bonan_ding/worldmem_data/minecraft/vae_features
14
+ - dataset.n_frames=1000
15
+ - +dataset.n_frames_valid=1100
16
+ - +dataset.customized_validation=true
17
+ - +dataset.wo_updown=false
18
+ - +dataset.angle_range=180
19
+ - +dataset.pos_range=8
20
+ - ++algorithm.n_tokens=4
21
+ - algorithm.x_shape=[16,18,32]
22
+ - ++algorithm.context_frames=100
23
+ - ++algorithm.log_video=true
24
+ - ++algorithm.diffusion.sampling_timesteps=20
25
+ - ++algorithm.dememwm.debug_force_all_streams=false
26
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
27
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
28
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
29
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
30
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
31
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
32
+ - ++algorithm.dememwm.anchor.enabled=true
33
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
34
+ - ++algorithm.dememwm.anchor.diverse_selection=true
35
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
36
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
37
+ - ++algorithm.dememwm.dynamic.enabled=true
38
+ - ++algorithm.dememwm.dynamic.recent_frames=4
39
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
40
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
41
+ - ++algorithm.dememwm.revisit.enabled=true
42
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
43
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
44
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
45
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
46
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
47
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
48
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
49
+ - ++algorithm.dememwm.revisit.max_frames=2
50
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
51
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
52
+ - ++algorithm.dememwm.cache.enabled=true
53
+ - ++algorithm.dememwm.cache.device=cpu
54
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
55
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
56
+ - ++algorithm.dememwm.cache.eviction_policy=none
57
+ - ++algorithm.dememwm.cache.no_evict=true
58
+ - ++algorithm.dememwm.cache.clear_between_videos=true
59
+ - ++algorithm.dememwm.cache.max_records=null
60
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
61
+ - ++algorithm.dememwm.curriculum.enabled=true
62
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
63
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
64
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
65
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
66
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
67
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
68
+ - experiment.training.batch_size=16
69
+ - experiment.training.optim.accumulate_grad_batches=1
70
+ - experiment.validation.batch_size=1
71
+ - experiment.validation.limit_batch=2
72
+ - experiment.training.checkpointing.every_n_train_steps=2000
73
+ - experiment.validation.val_every_n_step=2000
74
+ - experiment.training.max_steps=350000
outputs/2026-06-01/15-41-08/main.log ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ [2026-06-01 15:41:49,208][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP)
2
+ [2026-06-01 15:41:49,334][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True
3
+ [2026-06-01 15:41:49,334][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores
4
+ [2026-06-01 15:41:49,334][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs
5
+ [2026-06-01 15:41:49,334][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs
outputs/2026-06-03/20-22-54/.hydra/config.yaml ADDED
@@ -0,0 +1,234 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment:
2
+ debug: ${debug}
3
+ tasks:
4
+ - training
5
+ num_nodes: 1
6
+ training:
7
+ precision: 16-mixed
8
+ compile: false
9
+ lr: 2.0e-05
10
+ batch_size: 16
11
+ max_epochs: -1
12
+ max_steps: 350000
13
+ max_time: null
14
+ data:
15
+ num_workers: 4
16
+ shuffle: true
17
+ optim:
18
+ accumulate_grad_batches: 1
19
+ gradient_clip_val: 1.0
20
+ checkpointing:
21
+ every_n_train_steps: 2000
22
+ every_n_epochs: null
23
+ train_time_interval: null
24
+ save_last: true
25
+ enable_version_counter: false
26
+ validation:
27
+ precision: 16-mixed
28
+ compile: false
29
+ batch_size: 1
30
+ val_every_n_step: 2000
31
+ val_every_n_epoch: null
32
+ limit_batch: 2
33
+ inference_mode: true
34
+ data:
35
+ num_workers: 4
36
+ shuffle: false
37
+ test:
38
+ precision: 16-mixed
39
+ compile: false
40
+ batch_size: 1
41
+ limit_batch: 1
42
+ data:
43
+ num_workers: 4
44
+ shuffle: false
45
+ logging:
46
+ metrics: null
47
+ dataset:
48
+ debug: ${debug}
49
+ metadata: data/${dataset.name}/metadata.json
50
+ data_mean: 0.5
51
+ data_std: 0.5
52
+ save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
53
+ n_frames: 1000
54
+ context_length: 1
55
+ resolution: 128
56
+ observation_shape:
57
+ - 3
58
+ - ${dataset.resolution}
59
+ - ${dataset.resolution}
60
+ external_cond_dim: 0
61
+ validation_multiplier: 1
62
+ frame_skip: 1
63
+ action_cond_dim: 25
64
+ _name: video_minecraft_latent
65
+ precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
66
+ n_frames_valid: 1100
67
+ customized_validation: true
68
+ wo_updown: false
69
+ angle_range: 180
70
+ pos_range: 8
71
+ algorithm:
72
+ debug: ${debug}
73
+ lr: ${experiment.training.lr}
74
+ x_shape:
75
+ - 16
76
+ - 18
77
+ - 32
78
+ frame_stack: 1
79
+ frame_skip: ${dataset.frame_skip}
80
+ data_mean: ${dataset.data_mean}
81
+ data_std: ${dataset.data_std}
82
+ external_cond_dim: 0
83
+ context_frames: 100
84
+ weight_decay: 0.002
85
+ warmup_steps: 1000
86
+ optimizer_beta:
87
+ - 0.9
88
+ - 0.99
89
+ uncertainty_scale: 1
90
+ guidance_scale: 0.0
91
+ chunk_size: 1
92
+ scheduling_matrix: autoregressive
93
+ noise_level: random_all
94
+ causal: true
95
+ diffusion:
96
+ objective: pred_v
97
+ beta_schedule: sigmoid
98
+ schedule_fn_kwargs: {}
99
+ clip_noise: 20.0
100
+ use_snr: false
101
+ use_cum_snr: false
102
+ use_fused_snr: true
103
+ snr_clip: 5.0
104
+ cum_snr_decay: 0.96
105
+ timesteps: 1000
106
+ sampling_timesteps: 20
107
+ ddim_sampling_eta: 0.0
108
+ stabilization_level: 15
109
+ architecture:
110
+ network_size: 64
111
+ attn_heads: 4
112
+ attn_dim_head: 64
113
+ dim_mults:
114
+ - 1
115
+ - 2
116
+ - 4
117
+ - 8
118
+ resolution: ${dataset.resolution}
119
+ attn_resolutions:
120
+ - 16
121
+ - 32
122
+ - 64
123
+ - 128
124
+ use_init_temporal_attn: true
125
+ use_linear_attn: true
126
+ time_emb_type: rotary
127
+ n_frames: ${dataset.n_frames}
128
+ metadata: ${dataset.metadata}
129
+ action_cond_dim: 25
130
+ use_plucker: true
131
+ _name: dememwm_memory_dit
132
+ memory_token_cross_attention: true
133
+ memory_cross_attn_layers:
134
+ - 3
135
+ - 7
136
+ - 11
137
+ - 15
138
+ pose_cond_dim: 5
139
+ log_video: true
140
+ dememwm:
141
+ enabled: true
142
+ training_stage: stage_1
143
+ debug_force_all_streams: false
144
+ curriculum:
145
+ enabled: true
146
+ full_stage_start_step: 20000
147
+ freeze_vae: true
148
+ dit_freeze:
149
+ enabled: true
150
+ lr:
151
+ dememwm_modules: 4.0e-05
152
+ memory_adapters: 4.0e-05
153
+ full_dit: 1.0e-05
154
+ spatial_kernel_size: 3
155
+ anchor:
156
+ enabled: true
157
+ anchor_indices:
158
+ - 0
159
+ - 1
160
+ - 2
161
+ - 3
162
+ allow_generated_as_anchor: false
163
+ diverse_selection: true
164
+ compress:
165
+ downsample_ratio: 2
166
+ dynamic:
167
+ enabled: true
168
+ recent_frames: 4
169
+ conv_kernel_t: 3
170
+ conv_stride_t: 2
171
+ compress:
172
+ downsample_ratio: 2
173
+ revisit:
174
+ enabled: true
175
+ deterministic_pose_retrieval: true
176
+ fov_overlap_threshold: 0.6
177
+ plucker_weight: 0.1
178
+ max_frames: 2
179
+ fov_half_h: 52.5
180
+ fov_half_v: 37.5
181
+ fov_radius: 30.0
182
+ fov_yaw_samples: 25
183
+ fov_pitch_samples: 20
184
+ fov_depth_samples: 20
185
+ pose_preselect_topk: 64
186
+ plucker_grid_h: 4
187
+ plucker_grid_w: 4
188
+ plucker_focal_length: 0.35
189
+ compress:
190
+ downsample_ratio: 2
191
+ stage_policy:
192
+ noise_bucket_logging: true
193
+ eval_ablation:
194
+ enabled: false
195
+ branch: A_plus_D_plus_R_normal
196
+ generated_history_proxy:
197
+ enabled: true
198
+ start_step: 40000
199
+ ramp_steps: 40000
200
+ max_prob: 0.25
201
+ noise_std: 0.25
202
+ dropout_prob: 0.0
203
+ injection:
204
+ memory_hidden_size: 512
205
+ cache:
206
+ enabled: true
207
+ device: cpu
208
+ keep_raw_latents: all
209
+ keep_compressed_records: true
210
+ keep_prefix_anchors: true
211
+ eviction_policy: none
212
+ no_evict: true
213
+ clear_between_videos: true
214
+ max_records: null
215
+ on_capacity_exceeded: warn
216
+ checkpoint:
217
+ strict_dememwm_eval_load: true
218
+ n_tokens: 4
219
+ debug: false
220
+ wandb:
221
+ entity: turlin
222
+ project: worldmem
223
+ mode: offline
224
+ resume: null
225
+ load: null
226
+ auto_resume: true
227
+ resume_ckpt_path: null
228
+ name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
229
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
230
+ customized_load: true
231
+ seperate_load: true
232
+ diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
233
+ vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
234
+ only_tune_memory: false
outputs/2026-06-03/20-22-54/.hydra/hydra.yaml ADDED
@@ -0,0 +1,232 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
4
+ sweep:
5
+ dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
6
+ subdir: ${hydra.job.num}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ formatters:
71
+ simple:
72
+ format: '[%(asctime)s][HYDRA] %(message)s'
73
+ handlers:
74
+ console:
75
+ class: logging.StreamHandler
76
+ formatter: simple
77
+ stream: ext://sys.stdout
78
+ root:
79
+ level: INFO
80
+ handlers:
81
+ - console
82
+ loggers:
83
+ logging_example:
84
+ level: DEBUG
85
+ disable_existing_loggers: false
86
+ job_logging:
87
+ version: 1
88
+ formatters:
89
+ simple:
90
+ format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
91
+ handlers:
92
+ console:
93
+ class: logging.StreamHandler
94
+ formatter: simple
95
+ stream: ext://sys.stdout
96
+ file:
97
+ class: logging.FileHandler
98
+ formatter: simple
99
+ filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
100
+ root:
101
+ level: INFO
102
+ handlers:
103
+ - console
104
+ - file
105
+ disable_existing_loggers: false
106
+ env: {}
107
+ mode: RUN
108
+ searchpath: []
109
+ callbacks: {}
110
+ output_subdir: .hydra
111
+ overrides:
112
+ hydra:
113
+ - hydra.mode=RUN
114
+ task:
115
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
116
+ - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
117
+ - auto_resume=true
118
+ - experiment.tasks=[training]
119
+ - algorithm=dememwm_memory_dit
120
+ - +customized_load=true
121
+ - +seperate_load=true
122
+ - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
123
+ - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
124
+ - +only_tune_memory=false
125
+ - dataset=video_minecraft_latent
126
+ - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
127
+ - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
128
+ - dataset.n_frames=1000
129
+ - +dataset.n_frames_valid=1100
130
+ - +dataset.customized_validation=true
131
+ - +dataset.wo_updown=false
132
+ - +dataset.angle_range=180
133
+ - +dataset.pos_range=8
134
+ - ++algorithm.n_tokens=4
135
+ - algorithm.x_shape=[16,18,32]
136
+ - ++algorithm.context_frames=100
137
+ - ++algorithm.log_video=true
138
+ - ++algorithm.diffusion.sampling_timesteps=20
139
+ - ++algorithm.dememwm.debug_force_all_streams=false
140
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
141
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
142
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
143
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
144
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
145
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
146
+ - ++algorithm.dememwm.anchor.enabled=true
147
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
148
+ - ++algorithm.dememwm.anchor.diverse_selection=true
149
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
150
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
151
+ - ++algorithm.dememwm.dynamic.enabled=true
152
+ - ++algorithm.dememwm.dynamic.recent_frames=4
153
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
154
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
155
+ - ++algorithm.dememwm.revisit.enabled=true
156
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
157
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
158
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
159
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
160
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
161
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
162
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
163
+ - ++algorithm.dememwm.revisit.max_frames=2
164
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
165
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
166
+ - ++algorithm.dememwm.cache.enabled=true
167
+ - ++algorithm.dememwm.cache.device=cpu
168
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
169
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
170
+ - ++algorithm.dememwm.cache.eviction_policy=none
171
+ - ++algorithm.dememwm.cache.no_evict=true
172
+ - ++algorithm.dememwm.cache.clear_between_videos=true
173
+ - ++algorithm.dememwm.cache.max_records=null
174
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
175
+ - ++algorithm.dememwm.curriculum.enabled=true
176
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
177
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
178
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
179
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
180
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
181
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
182
+ - experiment.training.batch_size=16
183
+ - experiment.training.optim.accumulate_grad_batches=1
184
+ - experiment.validation.batch_size=1
185
+ - experiment.validation.limit_batch=2
186
+ - experiment.training.checkpointing.every_n_train_steps=2000
187
+ - experiment.validation.val_every_n_step=2000
188
+ - experiment.training.max_steps=350000
189
+ job:
190
+ name: main
191
+ chdir: null
192
+ 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
193
+ id: ???
194
+ num: ???
195
+ config_name: training
196
+ env_set: {}
197
+ env_copy: []
198
+ config:
199
+ override_dirname:
200
+ kv_sep: '='
201
+ item_sep: ','
202
+ exclude_keys: []
203
+ runtime:
204
+ version: 1.3.2
205
+ version_base: '1.3'
206
+ cwd: /proj/cvl/users/x_fahkh2/DeMemWM
207
+ config_sources:
208
+ - path: hydra.conf
209
+ schema: pkg
210
+ provider: hydra
211
+ - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations
212
+ schema: file
213
+ provider: main
214
+ - path: ''
215
+ schema: structured
216
+ provider: schema
217
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-03/20-22-54
218
+ choices:
219
+ cluster: null
220
+ algorithm: dememwm_memory_dit
221
+ dataset: video_minecraft_latent
222
+ experiment: exp_video
223
+ hydra/env: default
224
+ hydra/callbacks: null
225
+ hydra/job_logging: default
226
+ hydra/hydra_logging: default
227
+ hydra/hydra_help: default
228
+ hydra/help: default
229
+ hydra/sweeper: basic
230
+ hydra/launcher: basic
231
+ hydra/output: default
232
+ verbose: false
outputs/2026-06-03/20-22-54/.hydra/overrides.yaml ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
2
+ - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
3
+ - auto_resume=true
4
+ - experiment.tasks=[training]
5
+ - algorithm=dememwm_memory_dit
6
+ - +customized_load=true
7
+ - +seperate_load=true
8
+ - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
9
+ - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
10
+ - +only_tune_memory=false
11
+ - dataset=video_minecraft_latent
12
+ - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
13
+ - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
14
+ - dataset.n_frames=1000
15
+ - +dataset.n_frames_valid=1100
16
+ - +dataset.customized_validation=true
17
+ - +dataset.wo_updown=false
18
+ - +dataset.angle_range=180
19
+ - +dataset.pos_range=8
20
+ - ++algorithm.n_tokens=4
21
+ - algorithm.x_shape=[16,18,32]
22
+ - ++algorithm.context_frames=100
23
+ - ++algorithm.log_video=true
24
+ - ++algorithm.diffusion.sampling_timesteps=20
25
+ - ++algorithm.dememwm.debug_force_all_streams=false
26
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
27
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
28
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
29
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
30
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
31
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
32
+ - ++algorithm.dememwm.anchor.enabled=true
33
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
34
+ - ++algorithm.dememwm.anchor.diverse_selection=true
35
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
36
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
37
+ - ++algorithm.dememwm.dynamic.enabled=true
38
+ - ++algorithm.dememwm.dynamic.recent_frames=4
39
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
40
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
41
+ - ++algorithm.dememwm.revisit.enabled=true
42
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
43
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
44
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
45
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
46
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
47
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
48
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
49
+ - ++algorithm.dememwm.revisit.max_frames=2
50
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
51
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
52
+ - ++algorithm.dememwm.cache.enabled=true
53
+ - ++algorithm.dememwm.cache.device=cpu
54
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
55
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
56
+ - ++algorithm.dememwm.cache.eviction_policy=none
57
+ - ++algorithm.dememwm.cache.no_evict=true
58
+ - ++algorithm.dememwm.cache.clear_between_videos=true
59
+ - ++algorithm.dememwm.cache.max_records=null
60
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
61
+ - ++algorithm.dememwm.curriculum.enabled=true
62
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
63
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
64
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
65
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
66
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
67
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
68
+ - experiment.training.batch_size=16
69
+ - experiment.training.optim.accumulate_grad_batches=1
70
+ - experiment.validation.batch_size=1
71
+ - experiment.validation.limit_batch=2
72
+ - experiment.training.checkpointing.every_n_train_steps=2000
73
+ - experiment.validation.val_every_n_step=2000
74
+ - experiment.training.max_steps=350000
outputs/2026-06-03/20-22-54/main.log ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [2026-06-03 20:23:36,031][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP)
2
+ [2026-06-03 20:23:36,800][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True
3
+ [2026-06-03 20:23:36,800][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores
4
+ [2026-06-03 20:23:36,800][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs
5
+ [2026-06-03 20:23:36,800][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs
6
+ [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=''
7
+ [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=''
8
+ [2026-06-03 20:23:42,001][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8
9
+ [2026-06-03 20:23:42,005][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8
10
+ [2026-06-03 20:23:42,015][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8
11
+ [2026-06-03 20:23:42,026][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 1, MEMBER: 2/8
12
+ [2026-06-03 20:23:48,581][pytorch_lightning.utilities.rank_zero][INFO] - ----------------------------------------------------------------------------------------------------
13
+ distributed_backend=nccl
14
+ All distributed processes registered. Starting with 8 processes
15
+ ----------------------------------------------------------------------------------------------------
16
+
17
+ [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]
18
+ [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]
19
+ [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]
20
+ [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]
21
+ [2026-06-03 20:24:05,745][lightning.pytorch.callbacks.model_summary][INFO] -
22
+ | Name | Type | Params
23
+ -------------------------------------------------------------------------------------
24
+ 0 | diffusion_model | Diffusion | 631 M
25
+ 1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 887 K
26
+ 2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 296 K
27
+ 3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 296 K
28
+ 4 | dememwm_revisit_gate | RevisitRawGate | 4
29
+ 5 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M
30
+ 6 | vae | AutoencoderKL | 229 M
31
+ -------------------------------------------------------------------------------------
32
+ 632 M Trainable params
33
+ 231 M Non-trainable params
34
+ 864 M Total params
35
+ 3,456.995 Total estimated model params size (MB)
36
+ [2026-06-03 20:24:08,742][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
37
+ [2026-06-03 20:24:08,742][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
38
+ [2026-06-03 20:24:08,742][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
39
+ [2026-06-03 20:24:08,742][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
40
+ [2026-06-04 06:21:13,812][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15
41
+ [2026-06-04 06:21:13,813][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
42
+ [2026-06-04 06:21:13,812][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15
43
+ [2026-06-04 06:21:13,813][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
44
+ [2026-06-04 06:21:13,812][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 1] Received SIGTERM: 15
45
+ [2026-06-04 06:21:13,812][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15
46
+ [2026-06-04 06:21:13,814][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
47
+ [2026-06-04 06:21:13,814][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
outputs/2026-06-03/20-22-55/.hydra/config.yaml ADDED
@@ -0,0 +1,234 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment:
2
+ debug: ${debug}
3
+ tasks:
4
+ - training
5
+ num_nodes: 1
6
+ training:
7
+ precision: 16-mixed
8
+ compile: false
9
+ lr: 2.0e-05
10
+ batch_size: 16
11
+ max_epochs: -1
12
+ max_steps: 350000
13
+ max_time: null
14
+ data:
15
+ num_workers: 4
16
+ shuffle: true
17
+ optim:
18
+ accumulate_grad_batches: 1
19
+ gradient_clip_val: 1.0
20
+ checkpointing:
21
+ every_n_train_steps: 2000
22
+ every_n_epochs: null
23
+ train_time_interval: null
24
+ save_last: true
25
+ enable_version_counter: false
26
+ validation:
27
+ precision: 16-mixed
28
+ compile: false
29
+ batch_size: 1
30
+ val_every_n_step: 2000
31
+ val_every_n_epoch: null
32
+ limit_batch: 2
33
+ inference_mode: true
34
+ data:
35
+ num_workers: 4
36
+ shuffle: false
37
+ test:
38
+ precision: 16-mixed
39
+ compile: false
40
+ batch_size: 1
41
+ limit_batch: 1
42
+ data:
43
+ num_workers: 4
44
+ shuffle: false
45
+ logging:
46
+ metrics: null
47
+ dataset:
48
+ debug: ${debug}
49
+ metadata: data/${dataset.name}/metadata.json
50
+ data_mean: 0.5
51
+ data_std: 0.5
52
+ save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
53
+ n_frames: 1000
54
+ context_length: 1
55
+ resolution: 128
56
+ observation_shape:
57
+ - 3
58
+ - ${dataset.resolution}
59
+ - ${dataset.resolution}
60
+ external_cond_dim: 0
61
+ validation_multiplier: 1
62
+ frame_skip: 1
63
+ action_cond_dim: 25
64
+ _name: video_minecraft_latent
65
+ precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
66
+ n_frames_valid: 1100
67
+ customized_validation: true
68
+ wo_updown: false
69
+ angle_range: 180
70
+ pos_range: 8
71
+ algorithm:
72
+ debug: ${debug}
73
+ lr: ${experiment.training.lr}
74
+ x_shape:
75
+ - 16
76
+ - 18
77
+ - 32
78
+ frame_stack: 1
79
+ frame_skip: ${dataset.frame_skip}
80
+ data_mean: ${dataset.data_mean}
81
+ data_std: ${dataset.data_std}
82
+ external_cond_dim: 0
83
+ context_frames: 100
84
+ weight_decay: 0.002
85
+ warmup_steps: 1000
86
+ optimizer_beta:
87
+ - 0.9
88
+ - 0.99
89
+ uncertainty_scale: 1
90
+ guidance_scale: 0.0
91
+ chunk_size: 1
92
+ scheduling_matrix: autoregressive
93
+ noise_level: random_all
94
+ causal: true
95
+ diffusion:
96
+ objective: pred_v
97
+ beta_schedule: sigmoid
98
+ schedule_fn_kwargs: {}
99
+ clip_noise: 20.0
100
+ use_snr: false
101
+ use_cum_snr: false
102
+ use_fused_snr: true
103
+ snr_clip: 5.0
104
+ cum_snr_decay: 0.96
105
+ timesteps: 1000
106
+ sampling_timesteps: 20
107
+ ddim_sampling_eta: 0.0
108
+ stabilization_level: 15
109
+ architecture:
110
+ network_size: 64
111
+ attn_heads: 4
112
+ attn_dim_head: 64
113
+ dim_mults:
114
+ - 1
115
+ - 2
116
+ - 4
117
+ - 8
118
+ resolution: ${dataset.resolution}
119
+ attn_resolutions:
120
+ - 16
121
+ - 32
122
+ - 64
123
+ - 128
124
+ use_init_temporal_attn: true
125
+ use_linear_attn: true
126
+ time_emb_type: rotary
127
+ n_frames: ${dataset.n_frames}
128
+ metadata: ${dataset.metadata}
129
+ action_cond_dim: 25
130
+ use_plucker: true
131
+ _name: dememwm_memory_dit
132
+ memory_token_cross_attention: true
133
+ memory_cross_attn_layers:
134
+ - 3
135
+ - 7
136
+ - 11
137
+ - 15
138
+ pose_cond_dim: 5
139
+ log_video: true
140
+ dememwm:
141
+ enabled: true
142
+ training_stage: stage_1
143
+ debug_force_all_streams: false
144
+ curriculum:
145
+ enabled: true
146
+ full_stage_start_step: 20000
147
+ freeze_vae: true
148
+ dit_freeze:
149
+ enabled: true
150
+ lr:
151
+ dememwm_modules: 4.0e-05
152
+ memory_adapters: 4.0e-05
153
+ full_dit: 1.0e-05
154
+ spatial_kernel_size: 3
155
+ anchor:
156
+ enabled: true
157
+ anchor_indices:
158
+ - 0
159
+ - 1
160
+ - 2
161
+ - 3
162
+ allow_generated_as_anchor: false
163
+ diverse_selection: true
164
+ compress:
165
+ downsample_ratio: 2
166
+ dynamic:
167
+ enabled: true
168
+ recent_frames: 4
169
+ conv_kernel_t: 3
170
+ conv_stride_t: 2
171
+ compress:
172
+ downsample_ratio: 2
173
+ revisit:
174
+ enabled: true
175
+ deterministic_pose_retrieval: true
176
+ fov_overlap_threshold: 0.6
177
+ plucker_weight: 0.1
178
+ max_frames: 2
179
+ fov_half_h: 52.5
180
+ fov_half_v: 37.5
181
+ fov_radius: 30.0
182
+ fov_yaw_samples: 25
183
+ fov_pitch_samples: 20
184
+ fov_depth_samples: 20
185
+ pose_preselect_topk: 64
186
+ plucker_grid_h: 4
187
+ plucker_grid_w: 4
188
+ plucker_focal_length: 0.35
189
+ compress:
190
+ downsample_ratio: 2
191
+ stage_policy:
192
+ noise_bucket_logging: true
193
+ eval_ablation:
194
+ enabled: false
195
+ branch: A_plus_D_plus_R_normal
196
+ generated_history_proxy:
197
+ enabled: true
198
+ start_step: 40000
199
+ ramp_steps: 40000
200
+ max_prob: 0.25
201
+ noise_std: 0.25
202
+ dropout_prob: 0.0
203
+ injection:
204
+ memory_hidden_size: 512
205
+ cache:
206
+ enabled: true
207
+ device: cpu
208
+ keep_raw_latents: all
209
+ keep_compressed_records: true
210
+ keep_prefix_anchors: true
211
+ eviction_policy: none
212
+ no_evict: true
213
+ clear_between_videos: true
214
+ max_records: null
215
+ on_capacity_exceeded: warn
216
+ checkpoint:
217
+ strict_dememwm_eval_load: true
218
+ n_tokens: 4
219
+ debug: false
220
+ wandb:
221
+ entity: turlin
222
+ project: worldmem
223
+ mode: offline
224
+ resume: null
225
+ load: null
226
+ auto_resume: true
227
+ resume_ckpt_path: null
228
+ name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
229
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
230
+ customized_load: true
231
+ seperate_load: true
232
+ diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
233
+ vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
234
+ only_tune_memory: false
outputs/2026-06-03/20-22-55/.hydra/hydra.yaml ADDED
@@ -0,0 +1,232 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
4
+ sweep:
5
+ dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
6
+ subdir: ${hydra.job.num}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ formatters:
71
+ simple:
72
+ format: '[%(asctime)s][HYDRA] %(message)s'
73
+ handlers:
74
+ console:
75
+ class: logging.StreamHandler
76
+ formatter: simple
77
+ stream: ext://sys.stdout
78
+ root:
79
+ level: INFO
80
+ handlers:
81
+ - console
82
+ loggers:
83
+ logging_example:
84
+ level: DEBUG
85
+ disable_existing_loggers: false
86
+ job_logging:
87
+ version: 1
88
+ formatters:
89
+ simple:
90
+ format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
91
+ handlers:
92
+ console:
93
+ class: logging.StreamHandler
94
+ formatter: simple
95
+ stream: ext://sys.stdout
96
+ file:
97
+ class: logging.FileHandler
98
+ formatter: simple
99
+ filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
100
+ root:
101
+ level: INFO
102
+ handlers:
103
+ - console
104
+ - file
105
+ disable_existing_loggers: false
106
+ env: {}
107
+ mode: RUN
108
+ searchpath: []
109
+ callbacks: {}
110
+ output_subdir: .hydra
111
+ overrides:
112
+ hydra:
113
+ - hydra.mode=RUN
114
+ task:
115
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
116
+ - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
117
+ - auto_resume=true
118
+ - experiment.tasks=[training]
119
+ - algorithm=dememwm_memory_dit
120
+ - +customized_load=true
121
+ - +seperate_load=true
122
+ - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
123
+ - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
124
+ - +only_tune_memory=false
125
+ - dataset=video_minecraft_latent
126
+ - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
127
+ - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
128
+ - dataset.n_frames=1000
129
+ - +dataset.n_frames_valid=1100
130
+ - +dataset.customized_validation=true
131
+ - +dataset.wo_updown=false
132
+ - +dataset.angle_range=180
133
+ - +dataset.pos_range=8
134
+ - ++algorithm.n_tokens=4
135
+ - algorithm.x_shape=[16,18,32]
136
+ - ++algorithm.context_frames=100
137
+ - ++algorithm.log_video=true
138
+ - ++algorithm.diffusion.sampling_timesteps=20
139
+ - ++algorithm.dememwm.debug_force_all_streams=false
140
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
141
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
142
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
143
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
144
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
145
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
146
+ - ++algorithm.dememwm.anchor.enabled=true
147
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
148
+ - ++algorithm.dememwm.anchor.diverse_selection=true
149
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
150
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
151
+ - ++algorithm.dememwm.dynamic.enabled=true
152
+ - ++algorithm.dememwm.dynamic.recent_frames=4
153
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
154
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
155
+ - ++algorithm.dememwm.revisit.enabled=true
156
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
157
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
158
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
159
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
160
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
161
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
162
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
163
+ - ++algorithm.dememwm.revisit.max_frames=2
164
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
165
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
166
+ - ++algorithm.dememwm.cache.enabled=true
167
+ - ++algorithm.dememwm.cache.device=cpu
168
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
169
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
170
+ - ++algorithm.dememwm.cache.eviction_policy=none
171
+ - ++algorithm.dememwm.cache.no_evict=true
172
+ - ++algorithm.dememwm.cache.clear_between_videos=true
173
+ - ++algorithm.dememwm.cache.max_records=null
174
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
175
+ - ++algorithm.dememwm.curriculum.enabled=true
176
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
177
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
178
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
179
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
180
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
181
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
182
+ - experiment.training.batch_size=16
183
+ - experiment.training.optim.accumulate_grad_batches=1
184
+ - experiment.validation.batch_size=1
185
+ - experiment.validation.limit_batch=2
186
+ - experiment.training.checkpointing.every_n_train_steps=2000
187
+ - experiment.validation.val_every_n_step=2000
188
+ - experiment.training.max_steps=350000
189
+ job:
190
+ name: main
191
+ chdir: null
192
+ 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
193
+ id: ???
194
+ num: ???
195
+ config_name: training
196
+ env_set: {}
197
+ env_copy: []
198
+ config:
199
+ override_dirname:
200
+ kv_sep: '='
201
+ item_sep: ','
202
+ exclude_keys: []
203
+ runtime:
204
+ version: 1.3.2
205
+ version_base: '1.3'
206
+ cwd: /proj/cvl/users/x_fahkh2/DeMemWM
207
+ config_sources:
208
+ - path: hydra.conf
209
+ schema: pkg
210
+ provider: hydra
211
+ - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations
212
+ schema: file
213
+ provider: main
214
+ - path: ''
215
+ schema: structured
216
+ provider: schema
217
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-03/20-22-55
218
+ choices:
219
+ cluster: null
220
+ algorithm: dememwm_memory_dit
221
+ dataset: video_minecraft_latent
222
+ experiment: exp_video
223
+ hydra/env: default
224
+ hydra/callbacks: null
225
+ hydra/job_logging: default
226
+ hydra/hydra_logging: default
227
+ hydra/hydra_help: default
228
+ hydra/help: default
229
+ hydra/sweeper: basic
230
+ hydra/launcher: basic
231
+ hydra/output: default
232
+ verbose: false
outputs/2026-06-03/20-22-55/.hydra/overrides.yaml ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
2
+ - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
3
+ - auto_resume=true
4
+ - experiment.tasks=[training]
5
+ - algorithm=dememwm_memory_dit
6
+ - +customized_load=true
7
+ - +seperate_load=true
8
+ - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
9
+ - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
10
+ - +only_tune_memory=false
11
+ - dataset=video_minecraft_latent
12
+ - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
13
+ - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
14
+ - dataset.n_frames=1000
15
+ - +dataset.n_frames_valid=1100
16
+ - +dataset.customized_validation=true
17
+ - +dataset.wo_updown=false
18
+ - +dataset.angle_range=180
19
+ - +dataset.pos_range=8
20
+ - ++algorithm.n_tokens=4
21
+ - algorithm.x_shape=[16,18,32]
22
+ - ++algorithm.context_frames=100
23
+ - ++algorithm.log_video=true
24
+ - ++algorithm.diffusion.sampling_timesteps=20
25
+ - ++algorithm.dememwm.debug_force_all_streams=false
26
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
27
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
28
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
29
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
30
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
31
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
32
+ - ++algorithm.dememwm.anchor.enabled=true
33
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
34
+ - ++algorithm.dememwm.anchor.diverse_selection=true
35
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
36
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
37
+ - ++algorithm.dememwm.dynamic.enabled=true
38
+ - ++algorithm.dememwm.dynamic.recent_frames=4
39
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
40
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
41
+ - ++algorithm.dememwm.revisit.enabled=true
42
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
43
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
44
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
45
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
46
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
47
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
48
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
49
+ - ++algorithm.dememwm.revisit.max_frames=2
50
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
51
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
52
+ - ++algorithm.dememwm.cache.enabled=true
53
+ - ++algorithm.dememwm.cache.device=cpu
54
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
55
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
56
+ - ++algorithm.dememwm.cache.eviction_policy=none
57
+ - ++algorithm.dememwm.cache.no_evict=true
58
+ - ++algorithm.dememwm.cache.clear_between_videos=true
59
+ - ++algorithm.dememwm.cache.max_records=null
60
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
61
+ - ++algorithm.dememwm.curriculum.enabled=true
62
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
63
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
64
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
65
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
66
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
67
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
68
+ - experiment.training.batch_size=16
69
+ - experiment.training.optim.accumulate_grad_batches=1
70
+ - experiment.validation.batch_size=1
71
+ - experiment.validation.limit_batch=2
72
+ - experiment.training.checkpointing.every_n_train_steps=2000
73
+ - experiment.validation.val_every_n_step=2000
74
+ - experiment.training.max_steps=350000
outputs/2026-06-03/20-22-55/main.log ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ [2026-06-03 20:23:41,996][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8
2
+ [2026-06-03 20:23:42,011][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8
3
+ [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]
4
+ [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]
5
+ [2026-06-03 20:24:08,742][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
6
+ [2026-06-03 20:24:08,742][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
7
+ [2026-06-04 06:21:13,812][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15
8
+ [2026-06-04 06:21:13,813][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
9
+ [2026-06-04 06:21:13,812][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15
10
+ [2026-06-04 06:21:13,813][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
outputs/2026-06-03/20-22-56/.hydra/config.yaml ADDED
@@ -0,0 +1,234 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment:
2
+ debug: ${debug}
3
+ tasks:
4
+ - training
5
+ num_nodes: 1
6
+ training:
7
+ precision: 16-mixed
8
+ compile: false
9
+ lr: 2.0e-05
10
+ batch_size: 16
11
+ max_epochs: -1
12
+ max_steps: 350000
13
+ max_time: null
14
+ data:
15
+ num_workers: 4
16
+ shuffle: true
17
+ optim:
18
+ accumulate_grad_batches: 1
19
+ gradient_clip_val: 1.0
20
+ checkpointing:
21
+ every_n_train_steps: 2000
22
+ every_n_epochs: null
23
+ train_time_interval: null
24
+ save_last: true
25
+ enable_version_counter: false
26
+ validation:
27
+ precision: 16-mixed
28
+ compile: false
29
+ batch_size: 1
30
+ val_every_n_step: 2000
31
+ val_every_n_epoch: null
32
+ limit_batch: 2
33
+ inference_mode: true
34
+ data:
35
+ num_workers: 4
36
+ shuffle: false
37
+ test:
38
+ precision: 16-mixed
39
+ compile: false
40
+ batch_size: 1
41
+ limit_batch: 1
42
+ data:
43
+ num_workers: 4
44
+ shuffle: false
45
+ logging:
46
+ metrics: null
47
+ dataset:
48
+ debug: ${debug}
49
+ metadata: data/${dataset.name}/metadata.json
50
+ data_mean: 0.5
51
+ data_std: 0.5
52
+ save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
53
+ n_frames: 1000
54
+ context_length: 1
55
+ resolution: 128
56
+ observation_shape:
57
+ - 3
58
+ - ${dataset.resolution}
59
+ - ${dataset.resolution}
60
+ external_cond_dim: 0
61
+ validation_multiplier: 1
62
+ frame_skip: 1
63
+ action_cond_dim: 25
64
+ _name: video_minecraft_latent
65
+ precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
66
+ n_frames_valid: 1100
67
+ customized_validation: true
68
+ wo_updown: false
69
+ angle_range: 180
70
+ pos_range: 8
71
+ algorithm:
72
+ debug: ${debug}
73
+ lr: ${experiment.training.lr}
74
+ x_shape:
75
+ - 16
76
+ - 18
77
+ - 32
78
+ frame_stack: 1
79
+ frame_skip: ${dataset.frame_skip}
80
+ data_mean: ${dataset.data_mean}
81
+ data_std: ${dataset.data_std}
82
+ external_cond_dim: 0
83
+ context_frames: 100
84
+ weight_decay: 0.002
85
+ warmup_steps: 1000
86
+ optimizer_beta:
87
+ - 0.9
88
+ - 0.99
89
+ uncertainty_scale: 1
90
+ guidance_scale: 0.0
91
+ chunk_size: 1
92
+ scheduling_matrix: autoregressive
93
+ noise_level: random_all
94
+ causal: true
95
+ diffusion:
96
+ objective: pred_v
97
+ beta_schedule: sigmoid
98
+ schedule_fn_kwargs: {}
99
+ clip_noise: 20.0
100
+ use_snr: false
101
+ use_cum_snr: false
102
+ use_fused_snr: true
103
+ snr_clip: 5.0
104
+ cum_snr_decay: 0.96
105
+ timesteps: 1000
106
+ sampling_timesteps: 20
107
+ ddim_sampling_eta: 0.0
108
+ stabilization_level: 15
109
+ architecture:
110
+ network_size: 64
111
+ attn_heads: 4
112
+ attn_dim_head: 64
113
+ dim_mults:
114
+ - 1
115
+ - 2
116
+ - 4
117
+ - 8
118
+ resolution: ${dataset.resolution}
119
+ attn_resolutions:
120
+ - 16
121
+ - 32
122
+ - 64
123
+ - 128
124
+ use_init_temporal_attn: true
125
+ use_linear_attn: true
126
+ time_emb_type: rotary
127
+ n_frames: ${dataset.n_frames}
128
+ metadata: ${dataset.metadata}
129
+ action_cond_dim: 25
130
+ use_plucker: true
131
+ _name: dememwm_memory_dit
132
+ memory_token_cross_attention: true
133
+ memory_cross_attn_layers:
134
+ - 3
135
+ - 7
136
+ - 11
137
+ - 15
138
+ pose_cond_dim: 5
139
+ log_video: true
140
+ dememwm:
141
+ enabled: true
142
+ training_stage: stage_1
143
+ debug_force_all_streams: false
144
+ curriculum:
145
+ enabled: true
146
+ full_stage_start_step: 20000
147
+ freeze_vae: true
148
+ dit_freeze:
149
+ enabled: true
150
+ lr:
151
+ dememwm_modules: 4.0e-05
152
+ memory_adapters: 4.0e-05
153
+ full_dit: 1.0e-05
154
+ spatial_kernel_size: 3
155
+ anchor:
156
+ enabled: true
157
+ anchor_indices:
158
+ - 0
159
+ - 1
160
+ - 2
161
+ - 3
162
+ allow_generated_as_anchor: false
163
+ diverse_selection: true
164
+ compress:
165
+ downsample_ratio: 2
166
+ dynamic:
167
+ enabled: true
168
+ recent_frames: 4
169
+ conv_kernel_t: 3
170
+ conv_stride_t: 2
171
+ compress:
172
+ downsample_ratio: 2
173
+ revisit:
174
+ enabled: true
175
+ deterministic_pose_retrieval: true
176
+ fov_overlap_threshold: 0.6
177
+ plucker_weight: 0.1
178
+ max_frames: 2
179
+ fov_half_h: 52.5
180
+ fov_half_v: 37.5
181
+ fov_radius: 30.0
182
+ fov_yaw_samples: 25
183
+ fov_pitch_samples: 20
184
+ fov_depth_samples: 20
185
+ pose_preselect_topk: 64
186
+ plucker_grid_h: 4
187
+ plucker_grid_w: 4
188
+ plucker_focal_length: 0.35
189
+ compress:
190
+ downsample_ratio: 2
191
+ stage_policy:
192
+ noise_bucket_logging: true
193
+ eval_ablation:
194
+ enabled: false
195
+ branch: A_plus_D_plus_R_normal
196
+ generated_history_proxy:
197
+ enabled: true
198
+ start_step: 40000
199
+ ramp_steps: 40000
200
+ max_prob: 0.25
201
+ noise_std: 0.25
202
+ dropout_prob: 0.0
203
+ injection:
204
+ memory_hidden_size: 512
205
+ cache:
206
+ enabled: true
207
+ device: cpu
208
+ keep_raw_latents: all
209
+ keep_compressed_records: true
210
+ keep_prefix_anchors: true
211
+ eviction_policy: none
212
+ no_evict: true
213
+ clear_between_videos: true
214
+ max_records: null
215
+ on_capacity_exceeded: warn
216
+ checkpoint:
217
+ strict_dememwm_eval_load: true
218
+ n_tokens: 4
219
+ debug: false
220
+ wandb:
221
+ entity: turlin
222
+ project: worldmem
223
+ mode: offline
224
+ resume: null
225
+ load: null
226
+ auto_resume: true
227
+ resume_ckpt_path: null
228
+ name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
229
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
230
+ customized_load: true
231
+ seperate_load: true
232
+ diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
233
+ vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
234
+ only_tune_memory: false
outputs/2026-06-03/20-22-56/.hydra/hydra.yaml ADDED
@@ -0,0 +1,232 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
4
+ sweep:
5
+ dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
6
+ subdir: ${hydra.job.num}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ formatters:
71
+ simple:
72
+ format: '[%(asctime)s][HYDRA] %(message)s'
73
+ handlers:
74
+ console:
75
+ class: logging.StreamHandler
76
+ formatter: simple
77
+ stream: ext://sys.stdout
78
+ root:
79
+ level: INFO
80
+ handlers:
81
+ - console
82
+ loggers:
83
+ logging_example:
84
+ level: DEBUG
85
+ disable_existing_loggers: false
86
+ job_logging:
87
+ version: 1
88
+ formatters:
89
+ simple:
90
+ format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
91
+ handlers:
92
+ console:
93
+ class: logging.StreamHandler
94
+ formatter: simple
95
+ stream: ext://sys.stdout
96
+ file:
97
+ class: logging.FileHandler
98
+ formatter: simple
99
+ filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
100
+ root:
101
+ level: INFO
102
+ handlers:
103
+ - console
104
+ - file
105
+ disable_existing_loggers: false
106
+ env: {}
107
+ mode: RUN
108
+ searchpath: []
109
+ callbacks: {}
110
+ output_subdir: .hydra
111
+ overrides:
112
+ hydra:
113
+ - hydra.mode=RUN
114
+ task:
115
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
116
+ - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
117
+ - auto_resume=true
118
+ - experiment.tasks=[training]
119
+ - algorithm=dememwm_memory_dit
120
+ - +customized_load=true
121
+ - +seperate_load=true
122
+ - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
123
+ - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
124
+ - +only_tune_memory=false
125
+ - dataset=video_minecraft_latent
126
+ - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
127
+ - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
128
+ - dataset.n_frames=1000
129
+ - +dataset.n_frames_valid=1100
130
+ - +dataset.customized_validation=true
131
+ - +dataset.wo_updown=false
132
+ - +dataset.angle_range=180
133
+ - +dataset.pos_range=8
134
+ - ++algorithm.n_tokens=4
135
+ - algorithm.x_shape=[16,18,32]
136
+ - ++algorithm.context_frames=100
137
+ - ++algorithm.log_video=true
138
+ - ++algorithm.diffusion.sampling_timesteps=20
139
+ - ++algorithm.dememwm.debug_force_all_streams=false
140
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
141
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
142
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
143
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
144
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
145
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
146
+ - ++algorithm.dememwm.anchor.enabled=true
147
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
148
+ - ++algorithm.dememwm.anchor.diverse_selection=true
149
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
150
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
151
+ - ++algorithm.dememwm.dynamic.enabled=true
152
+ - ++algorithm.dememwm.dynamic.recent_frames=4
153
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
154
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
155
+ - ++algorithm.dememwm.revisit.enabled=true
156
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
157
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
158
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
159
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
160
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
161
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
162
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
163
+ - ++algorithm.dememwm.revisit.max_frames=2
164
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
165
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
166
+ - ++algorithm.dememwm.cache.enabled=true
167
+ - ++algorithm.dememwm.cache.device=cpu
168
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
169
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
170
+ - ++algorithm.dememwm.cache.eviction_policy=none
171
+ - ++algorithm.dememwm.cache.no_evict=true
172
+ - ++algorithm.dememwm.cache.clear_between_videos=true
173
+ - ++algorithm.dememwm.cache.max_records=null
174
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
175
+ - ++algorithm.dememwm.curriculum.enabled=true
176
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
177
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
178
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
179
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
180
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
181
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
182
+ - experiment.training.batch_size=16
183
+ - experiment.training.optim.accumulate_grad_batches=1
184
+ - experiment.validation.batch_size=1
185
+ - experiment.validation.limit_batch=2
186
+ - experiment.training.checkpointing.every_n_train_steps=2000
187
+ - experiment.validation.val_every_n_step=2000
188
+ - experiment.training.max_steps=350000
189
+ job:
190
+ name: main
191
+ chdir: null
192
+ 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
193
+ id: ???
194
+ num: ???
195
+ config_name: training
196
+ env_set: {}
197
+ env_copy: []
198
+ config:
199
+ override_dirname:
200
+ kv_sep: '='
201
+ item_sep: ','
202
+ exclude_keys: []
203
+ runtime:
204
+ version: 1.3.2
205
+ version_base: '1.3'
206
+ cwd: /proj/cvl/users/x_fahkh2/DeMemWM
207
+ config_sources:
208
+ - path: hydra.conf
209
+ schema: pkg
210
+ provider: hydra
211
+ - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations
212
+ schema: file
213
+ provider: main
214
+ - path: ''
215
+ schema: structured
216
+ provider: schema
217
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-03/20-22-56
218
+ choices:
219
+ cluster: null
220
+ algorithm: dememwm_memory_dit
221
+ dataset: video_minecraft_latent
222
+ experiment: exp_video
223
+ hydra/env: default
224
+ hydra/callbacks: null
225
+ hydra/job_logging: default
226
+ hydra/hydra_logging: default
227
+ hydra/hydra_help: default
228
+ hydra/help: default
229
+ hydra/sweeper: basic
230
+ hydra/launcher: basic
231
+ hydra/output: default
232
+ verbose: false
outputs/2026-06-03/20-22-56/.hydra/overrides.yaml ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
2
+ - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
3
+ - auto_resume=true
4
+ - experiment.tasks=[training]
5
+ - algorithm=dememwm_memory_dit
6
+ - +customized_load=true
7
+ - +seperate_load=true
8
+ - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
9
+ - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
10
+ - +only_tune_memory=false
11
+ - dataset=video_minecraft_latent
12
+ - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
13
+ - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
14
+ - dataset.n_frames=1000
15
+ - +dataset.n_frames_valid=1100
16
+ - +dataset.customized_validation=true
17
+ - +dataset.wo_updown=false
18
+ - +dataset.angle_range=180
19
+ - +dataset.pos_range=8
20
+ - ++algorithm.n_tokens=4
21
+ - algorithm.x_shape=[16,18,32]
22
+ - ++algorithm.context_frames=100
23
+ - ++algorithm.log_video=true
24
+ - ++algorithm.diffusion.sampling_timesteps=20
25
+ - ++algorithm.dememwm.debug_force_all_streams=false
26
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
27
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
28
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
29
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
30
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
31
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
32
+ - ++algorithm.dememwm.anchor.enabled=true
33
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
34
+ - ++algorithm.dememwm.anchor.diverse_selection=true
35
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
36
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
37
+ - ++algorithm.dememwm.dynamic.enabled=true
38
+ - ++algorithm.dememwm.dynamic.recent_frames=4
39
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
40
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
41
+ - ++algorithm.dememwm.revisit.enabled=true
42
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
43
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
44
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
45
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
46
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
47
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
48
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
49
+ - ++algorithm.dememwm.revisit.max_frames=2
50
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
51
+ - ++algorithm.dememwm.stage_policy.noise_bucket_logging=true
52
+ - ++algorithm.dememwm.cache.enabled=true
53
+ - ++algorithm.dememwm.cache.device=cpu
54
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
55
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
56
+ - ++algorithm.dememwm.cache.eviction_policy=none
57
+ - ++algorithm.dememwm.cache.no_evict=true
58
+ - ++algorithm.dememwm.cache.clear_between_videos=true
59
+ - ++algorithm.dememwm.cache.max_records=null
60
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
61
+ - ++algorithm.dememwm.curriculum.enabled=true
62
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
63
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
64
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
65
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
66
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
67
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
68
+ - experiment.training.batch_size=16
69
+ - experiment.training.optim.accumulate_grad_batches=1
70
+ - experiment.validation.batch_size=1
71
+ - experiment.validation.limit_batch=2
72
+ - experiment.training.checkpointing.every_n_train_steps=2000
73
+ - experiment.validation.val_every_n_step=2000
74
+ - experiment.training.max_steps=350000
outputs/2026-06-03/20-22-56/main.log ADDED
@@ -0,0 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
1
+ [2026-06-03 20:23:42,010][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8
2
+ [2026-06-03 20:23:42,017][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8
3
+ [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]
4
+ [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]
5
+ [2026-06-03 20:24:08,742][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
6
+ [2026-06-03 20:24:08,742][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
7
+ [2026-06-04 06:21:13,812][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15
8
+ [2026-06-04 06:21:13,813][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
9
+ [2026-06-04 06:21:13,813][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15
10
+ [2026-06-04 06:21:13,813][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
outputs/2026-06-04/06-22-27/.hydra/config.yaml ADDED
@@ -0,0 +1,231 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment:
2
+ debug: ${debug}
3
+ tasks:
4
+ - training
5
+ num_nodes: 1
6
+ training:
7
+ precision: 16-mixed
8
+ compile: false
9
+ lr: 2.0e-05
10
+ batch_size: 16
11
+ max_epochs: -1
12
+ max_steps: 350000
13
+ max_time: null
14
+ data:
15
+ num_workers: 4
16
+ shuffle: true
17
+ optim:
18
+ accumulate_grad_batches: 1
19
+ gradient_clip_val: 1.0
20
+ checkpointing:
21
+ every_n_train_steps: 5000
22
+ every_n_epochs: null
23
+ train_time_interval: null
24
+ save_last: true
25
+ enable_version_counter: false
26
+ validation:
27
+ precision: 16-mixed
28
+ compile: false
29
+ batch_size: 1
30
+ val_every_n_step: 5000
31
+ val_every_n_epoch: null
32
+ limit_batch: 1
33
+ inference_mode: true
34
+ data:
35
+ num_workers: 4
36
+ shuffle: false
37
+ test:
38
+ precision: 16-mixed
39
+ compile: false
40
+ batch_size: 1
41
+ limit_batch: 1
42
+ data:
43
+ num_workers: 4
44
+ shuffle: false
45
+ logging:
46
+ metrics: null
47
+ dataset:
48
+ debug: ${debug}
49
+ metadata: data/${dataset.name}/metadata.json
50
+ data_mean: 0.5
51
+ data_std: 0.5
52
+ save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
53
+ n_frames: 1000
54
+ context_length: 1
55
+ resolution: 128
56
+ observation_shape:
57
+ - 3
58
+ - ${dataset.resolution}
59
+ - ${dataset.resolution}
60
+ external_cond_dim: 0
61
+ validation_multiplier: 1
62
+ frame_skip: 1
63
+ action_cond_dim: 25
64
+ _name: video_minecraft_latent
65
+ precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
66
+ n_frames_valid: 500
67
+ customized_validation: true
68
+ wo_updown: false
69
+ angle_range: 180
70
+ pos_range: 8
71
+ algorithm:
72
+ debug: ${debug}
73
+ lr: ${experiment.training.lr}
74
+ x_shape:
75
+ - 16
76
+ - 18
77
+ - 32
78
+ frame_stack: 1
79
+ frame_skip: ${dataset.frame_skip}
80
+ data_mean: ${dataset.data_mean}
81
+ data_std: ${dataset.data_std}
82
+ external_cond_dim: 0
83
+ context_frames: 100
84
+ weight_decay: 0.002
85
+ warmup_steps: 1000
86
+ optimizer_beta:
87
+ - 0.9
88
+ - 0.99
89
+ uncertainty_scale: 1
90
+ guidance_scale: 0.0
91
+ chunk_size: 1
92
+ scheduling_matrix: autoregressive
93
+ noise_level: random_all
94
+ causal: true
95
+ diffusion:
96
+ objective: pred_v
97
+ beta_schedule: sigmoid
98
+ schedule_fn_kwargs: {}
99
+ clip_noise: 20.0
100
+ use_snr: false
101
+ use_cum_snr: false
102
+ use_fused_snr: true
103
+ snr_clip: 5.0
104
+ cum_snr_decay: 0.96
105
+ timesteps: 1000
106
+ sampling_timesteps: 20
107
+ ddim_sampling_eta: 0.0
108
+ stabilization_level: 15
109
+ architecture:
110
+ network_size: 64
111
+ attn_heads: 4
112
+ attn_dim_head: 64
113
+ dim_mults:
114
+ - 1
115
+ - 2
116
+ - 4
117
+ - 8
118
+ resolution: ${dataset.resolution}
119
+ attn_resolutions:
120
+ - 16
121
+ - 32
122
+ - 64
123
+ - 128
124
+ use_init_temporal_attn: true
125
+ use_linear_attn: true
126
+ time_emb_type: rotary
127
+ n_frames: ${dataset.n_frames}
128
+ metadata: ${dataset.metadata}
129
+ action_cond_dim: 25
130
+ use_plucker: true
131
+ _name: dememwm_memory_dit
132
+ memory_token_cross_attention: true
133
+ memory_cross_attn_layers:
134
+ - 3
135
+ - 7
136
+ - 11
137
+ - 15
138
+ pose_cond_dim: 5
139
+ log_video: true
140
+ dememwm:
141
+ enabled: true
142
+ training_stage: stage_1
143
+ curriculum:
144
+ enabled: true
145
+ full_stage_start_step: 20000
146
+ freeze_vae: true
147
+ dit_freeze:
148
+ enabled: true
149
+ lr:
150
+ dememwm_modules: 4.0e-05
151
+ memory_adapters: 4.0e-05
152
+ full_dit: 1.0e-05
153
+ spatial_kernel_size: 3
154
+ anchor:
155
+ enabled: true
156
+ anchor_indices:
157
+ - 0
158
+ - 1
159
+ - 2
160
+ - 3
161
+ allow_generated_as_anchor: false
162
+ diverse_selection: true
163
+ compress:
164
+ downsample_ratio: 2
165
+ dynamic:
166
+ enabled: true
167
+ recent_frames: 4
168
+ conv_kernel_t: 3
169
+ conv_stride_t: 2
170
+ compress:
171
+ downsample_ratio: 2
172
+ revisit:
173
+ enabled: true
174
+ deterministic_pose_retrieval: true
175
+ fov_overlap_threshold: 0.6
176
+ plucker_weight: 0.1
177
+ max_frames: 2
178
+ fov_half_h: 52.5
179
+ fov_half_v: 37.5
180
+ fov_radius: 30.0
181
+ fov_yaw_samples: 25
182
+ fov_pitch_samples: 20
183
+ fov_depth_samples: 20
184
+ pose_preselect_topk: 64
185
+ plucker_grid_h: 4
186
+ plucker_grid_w: 4
187
+ plucker_focal_length: 0.35
188
+ compress:
189
+ downsample_ratio: 2
190
+ eval_ablation:
191
+ enabled: false
192
+ branch: A_plus_D_plus_R_normal
193
+ generated_history_proxy:
194
+ enabled: true
195
+ start_step: 40000
196
+ ramp_steps: 40000
197
+ max_prob: 0.25
198
+ noise_std: 0.25
199
+ dropout_prob: 0.0
200
+ injection:
201
+ memory_hidden_size: 512
202
+ cache:
203
+ enabled: true
204
+ device: cpu
205
+ keep_raw_latents: all
206
+ keep_compressed_records: true
207
+ keep_prefix_anchors: true
208
+ eviction_policy: none
209
+ no_evict: true
210
+ clear_between_videos: true
211
+ max_records: null
212
+ on_capacity_exceeded: warn
213
+ checkpoint:
214
+ strict_dememwm_eval_load: true
215
+ n_tokens: 4
216
+ debug: false
217
+ wandb:
218
+ entity: turlin
219
+ project: worldmem
220
+ mode: offline
221
+ resume: null
222
+ load: null
223
+ auto_resume: true
224
+ resume_ckpt_path: null
225
+ name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
226
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
227
+ customized_load: true
228
+ seperate_load: true
229
+ diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
230
+ vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
231
+ only_tune_memory: false
outputs/2026-06-04/06-22-27/.hydra/hydra.yaml ADDED
@@ -0,0 +1,230 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
4
+ sweep:
5
+ dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
6
+ subdir: ${hydra.job.num}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ formatters:
71
+ simple:
72
+ format: '[%(asctime)s][HYDRA] %(message)s'
73
+ handlers:
74
+ console:
75
+ class: logging.StreamHandler
76
+ formatter: simple
77
+ stream: ext://sys.stdout
78
+ root:
79
+ level: INFO
80
+ handlers:
81
+ - console
82
+ loggers:
83
+ logging_example:
84
+ level: DEBUG
85
+ disable_existing_loggers: false
86
+ job_logging:
87
+ version: 1
88
+ formatters:
89
+ simple:
90
+ format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
91
+ handlers:
92
+ console:
93
+ class: logging.StreamHandler
94
+ formatter: simple
95
+ stream: ext://sys.stdout
96
+ file:
97
+ class: logging.FileHandler
98
+ formatter: simple
99
+ filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
100
+ root:
101
+ level: INFO
102
+ handlers:
103
+ - console
104
+ - file
105
+ disable_existing_loggers: false
106
+ env: {}
107
+ mode: RUN
108
+ searchpath: []
109
+ callbacks: {}
110
+ output_subdir: .hydra
111
+ overrides:
112
+ hydra:
113
+ - hydra.mode=RUN
114
+ task:
115
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
116
+ - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
117
+ - auto_resume=true
118
+ - experiment.tasks=[training]
119
+ - algorithm=dememwm_memory_dit
120
+ - +customized_load=true
121
+ - +seperate_load=true
122
+ - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
123
+ - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
124
+ - +only_tune_memory=false
125
+ - dataset=video_minecraft_latent
126
+ - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
127
+ - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
128
+ - dataset.n_frames=1000
129
+ - +dataset.n_frames_valid=500
130
+ - +dataset.customized_validation=true
131
+ - +dataset.wo_updown=false
132
+ - +dataset.angle_range=180
133
+ - +dataset.pos_range=8
134
+ - ++algorithm.n_tokens=4
135
+ - algorithm.x_shape=[16,18,32]
136
+ - ++algorithm.context_frames=100
137
+ - ++algorithm.log_video=true
138
+ - ++algorithm.diffusion.sampling_timesteps=20
139
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
140
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
141
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
142
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
143
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
144
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
145
+ - ++algorithm.dememwm.anchor.enabled=true
146
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
147
+ - ++algorithm.dememwm.anchor.diverse_selection=true
148
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
149
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
150
+ - ++algorithm.dememwm.dynamic.enabled=true
151
+ - ++algorithm.dememwm.dynamic.recent_frames=4
152
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
153
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
154
+ - ++algorithm.dememwm.revisit.enabled=true
155
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
156
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
157
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
158
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
159
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
160
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
161
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
162
+ - ++algorithm.dememwm.revisit.max_frames=2
163
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
164
+ - ++algorithm.dememwm.cache.enabled=true
165
+ - ++algorithm.dememwm.cache.device=cpu
166
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
167
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
168
+ - ++algorithm.dememwm.cache.eviction_policy=none
169
+ - ++algorithm.dememwm.cache.no_evict=true
170
+ - ++algorithm.dememwm.cache.clear_between_videos=true
171
+ - ++algorithm.dememwm.cache.max_records=null
172
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
173
+ - ++algorithm.dememwm.curriculum.enabled=true
174
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
175
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
176
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
177
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
178
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
179
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
180
+ - experiment.training.batch_size=16
181
+ - experiment.training.optim.accumulate_grad_batches=1
182
+ - experiment.validation.batch_size=1
183
+ - experiment.validation.limit_batch=1
184
+ - experiment.training.checkpointing.every_n_train_steps=5000
185
+ - experiment.validation.val_every_n_step=5000
186
+ - experiment.training.max_steps=350000
187
+ job:
188
+ name: main
189
+ chdir: null
190
+ 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
191
+ id: ???
192
+ num: ???
193
+ config_name: training
194
+ env_set: {}
195
+ env_copy: []
196
+ config:
197
+ override_dirname:
198
+ kv_sep: '='
199
+ item_sep: ','
200
+ exclude_keys: []
201
+ runtime:
202
+ version: 1.3.2
203
+ version_base: '1.3'
204
+ cwd: /proj/cvl/users/x_fahkh2/DeMemWM
205
+ config_sources:
206
+ - path: hydra.conf
207
+ schema: pkg
208
+ provider: hydra
209
+ - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations
210
+ schema: file
211
+ provider: main
212
+ - path: ''
213
+ schema: structured
214
+ provider: schema
215
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-04/06-22-27
216
+ choices:
217
+ cluster: null
218
+ algorithm: dememwm_memory_dit
219
+ dataset: video_minecraft_latent
220
+ experiment: exp_video
221
+ hydra/env: default
222
+ hydra/callbacks: null
223
+ hydra/job_logging: default
224
+ hydra/hydra_logging: default
225
+ hydra/hydra_help: default
226
+ hydra/help: default
227
+ hydra/sweeper: basic
228
+ hydra/launcher: basic
229
+ hydra/output: default
230
+ verbose: false
outputs/2026-06-04/06-22-27/.hydra/overrides.yaml ADDED
@@ -0,0 +1,72 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
2
+ - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
3
+ - auto_resume=true
4
+ - experiment.tasks=[training]
5
+ - algorithm=dememwm_memory_dit
6
+ - +customized_load=true
7
+ - +seperate_load=true
8
+ - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
9
+ - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
10
+ - +only_tune_memory=false
11
+ - dataset=video_minecraft_latent
12
+ - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
13
+ - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
14
+ - dataset.n_frames=1000
15
+ - +dataset.n_frames_valid=500
16
+ - +dataset.customized_validation=true
17
+ - +dataset.wo_updown=false
18
+ - +dataset.angle_range=180
19
+ - +dataset.pos_range=8
20
+ - ++algorithm.n_tokens=4
21
+ - algorithm.x_shape=[16,18,32]
22
+ - ++algorithm.context_frames=100
23
+ - ++algorithm.log_video=true
24
+ - ++algorithm.diffusion.sampling_timesteps=20
25
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
26
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
27
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
28
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
29
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
30
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
31
+ - ++algorithm.dememwm.anchor.enabled=true
32
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
33
+ - ++algorithm.dememwm.anchor.diverse_selection=true
34
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
35
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
36
+ - ++algorithm.dememwm.dynamic.enabled=true
37
+ - ++algorithm.dememwm.dynamic.recent_frames=4
38
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
39
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
40
+ - ++algorithm.dememwm.revisit.enabled=true
41
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
42
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
43
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
44
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
45
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
46
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
47
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
48
+ - ++algorithm.dememwm.revisit.max_frames=2
49
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
50
+ - ++algorithm.dememwm.cache.enabled=true
51
+ - ++algorithm.dememwm.cache.device=cpu
52
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
53
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
54
+ - ++algorithm.dememwm.cache.eviction_policy=none
55
+ - ++algorithm.dememwm.cache.no_evict=true
56
+ - ++algorithm.dememwm.cache.clear_between_videos=true
57
+ - ++algorithm.dememwm.cache.max_records=null
58
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
59
+ - ++algorithm.dememwm.curriculum.enabled=true
60
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
61
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
62
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
63
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
64
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
65
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
66
+ - experiment.training.batch_size=16
67
+ - experiment.training.optim.accumulate_grad_batches=1
68
+ - experiment.validation.batch_size=1
69
+ - experiment.validation.limit_batch=1
70
+ - experiment.training.checkpointing.every_n_train_steps=5000
71
+ - experiment.validation.val_every_n_step=5000
72
+ - experiment.training.max_steps=350000
outputs/2026-06-04/06-22-27/main.log ADDED
File without changes
outputs/2026-06-04/06-22-28/.hydra/config.yaml ADDED
@@ -0,0 +1,231 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment:
2
+ debug: ${debug}
3
+ tasks:
4
+ - training
5
+ num_nodes: 1
6
+ training:
7
+ precision: 16-mixed
8
+ compile: false
9
+ lr: 2.0e-05
10
+ batch_size: 16
11
+ max_epochs: -1
12
+ max_steps: 350000
13
+ max_time: null
14
+ data:
15
+ num_workers: 4
16
+ shuffle: true
17
+ optim:
18
+ accumulate_grad_batches: 1
19
+ gradient_clip_val: 1.0
20
+ checkpointing:
21
+ every_n_train_steps: 5000
22
+ every_n_epochs: null
23
+ train_time_interval: null
24
+ save_last: true
25
+ enable_version_counter: false
26
+ validation:
27
+ precision: 16-mixed
28
+ compile: false
29
+ batch_size: 1
30
+ val_every_n_step: 5000
31
+ val_every_n_epoch: null
32
+ limit_batch: 1
33
+ inference_mode: true
34
+ data:
35
+ num_workers: 4
36
+ shuffle: false
37
+ test:
38
+ precision: 16-mixed
39
+ compile: false
40
+ batch_size: 1
41
+ limit_batch: 1
42
+ data:
43
+ num_workers: 4
44
+ shuffle: false
45
+ logging:
46
+ metrics: null
47
+ dataset:
48
+ debug: ${debug}
49
+ metadata: data/${dataset.name}/metadata.json
50
+ data_mean: 0.5
51
+ data_std: 0.5
52
+ save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
53
+ n_frames: 1000
54
+ context_length: 1
55
+ resolution: 128
56
+ observation_shape:
57
+ - 3
58
+ - ${dataset.resolution}
59
+ - ${dataset.resolution}
60
+ external_cond_dim: 0
61
+ validation_multiplier: 1
62
+ frame_skip: 1
63
+ action_cond_dim: 25
64
+ _name: video_minecraft_latent
65
+ precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
66
+ n_frames_valid: 500
67
+ customized_validation: true
68
+ wo_updown: false
69
+ angle_range: 180
70
+ pos_range: 8
71
+ algorithm:
72
+ debug: ${debug}
73
+ lr: ${experiment.training.lr}
74
+ x_shape:
75
+ - 16
76
+ - 18
77
+ - 32
78
+ frame_stack: 1
79
+ frame_skip: ${dataset.frame_skip}
80
+ data_mean: ${dataset.data_mean}
81
+ data_std: ${dataset.data_std}
82
+ external_cond_dim: 0
83
+ context_frames: 100
84
+ weight_decay: 0.002
85
+ warmup_steps: 1000
86
+ optimizer_beta:
87
+ - 0.9
88
+ - 0.99
89
+ uncertainty_scale: 1
90
+ guidance_scale: 0.0
91
+ chunk_size: 1
92
+ scheduling_matrix: autoregressive
93
+ noise_level: random_all
94
+ causal: true
95
+ diffusion:
96
+ objective: pred_v
97
+ beta_schedule: sigmoid
98
+ schedule_fn_kwargs: {}
99
+ clip_noise: 20.0
100
+ use_snr: false
101
+ use_cum_snr: false
102
+ use_fused_snr: true
103
+ snr_clip: 5.0
104
+ cum_snr_decay: 0.96
105
+ timesteps: 1000
106
+ sampling_timesteps: 20
107
+ ddim_sampling_eta: 0.0
108
+ stabilization_level: 15
109
+ architecture:
110
+ network_size: 64
111
+ attn_heads: 4
112
+ attn_dim_head: 64
113
+ dim_mults:
114
+ - 1
115
+ - 2
116
+ - 4
117
+ - 8
118
+ resolution: ${dataset.resolution}
119
+ attn_resolutions:
120
+ - 16
121
+ - 32
122
+ - 64
123
+ - 128
124
+ use_init_temporal_attn: true
125
+ use_linear_attn: true
126
+ time_emb_type: rotary
127
+ n_frames: ${dataset.n_frames}
128
+ metadata: ${dataset.metadata}
129
+ action_cond_dim: 25
130
+ use_plucker: true
131
+ _name: dememwm_memory_dit
132
+ memory_token_cross_attention: true
133
+ memory_cross_attn_layers:
134
+ - 3
135
+ - 7
136
+ - 11
137
+ - 15
138
+ pose_cond_dim: 5
139
+ log_video: true
140
+ dememwm:
141
+ enabled: true
142
+ training_stage: stage_1
143
+ curriculum:
144
+ enabled: true
145
+ full_stage_start_step: 20000
146
+ freeze_vae: true
147
+ dit_freeze:
148
+ enabled: true
149
+ lr:
150
+ dememwm_modules: 4.0e-05
151
+ memory_adapters: 4.0e-05
152
+ full_dit: 1.0e-05
153
+ spatial_kernel_size: 3
154
+ anchor:
155
+ enabled: true
156
+ anchor_indices:
157
+ - 0
158
+ - 1
159
+ - 2
160
+ - 3
161
+ allow_generated_as_anchor: false
162
+ diverse_selection: true
163
+ compress:
164
+ downsample_ratio: 2
165
+ dynamic:
166
+ enabled: true
167
+ recent_frames: 4
168
+ conv_kernel_t: 3
169
+ conv_stride_t: 2
170
+ compress:
171
+ downsample_ratio: 2
172
+ revisit:
173
+ enabled: true
174
+ deterministic_pose_retrieval: true
175
+ fov_overlap_threshold: 0.6
176
+ plucker_weight: 0.1
177
+ max_frames: 2
178
+ fov_half_h: 52.5
179
+ fov_half_v: 37.5
180
+ fov_radius: 30.0
181
+ fov_yaw_samples: 25
182
+ fov_pitch_samples: 20
183
+ fov_depth_samples: 20
184
+ pose_preselect_topk: 64
185
+ plucker_grid_h: 4
186
+ plucker_grid_w: 4
187
+ plucker_focal_length: 0.35
188
+ compress:
189
+ downsample_ratio: 2
190
+ eval_ablation:
191
+ enabled: false
192
+ branch: A_plus_D_plus_R_normal
193
+ generated_history_proxy:
194
+ enabled: true
195
+ start_step: 40000
196
+ ramp_steps: 40000
197
+ max_prob: 0.25
198
+ noise_std: 0.25
199
+ dropout_prob: 0.0
200
+ injection:
201
+ memory_hidden_size: 512
202
+ cache:
203
+ enabled: true
204
+ device: cpu
205
+ keep_raw_latents: all
206
+ keep_compressed_records: true
207
+ keep_prefix_anchors: true
208
+ eviction_policy: none
209
+ no_evict: true
210
+ clear_between_videos: true
211
+ max_records: null
212
+ on_capacity_exceeded: warn
213
+ checkpoint:
214
+ strict_dememwm_eval_load: true
215
+ n_tokens: 4
216
+ debug: false
217
+ wandb:
218
+ entity: turlin
219
+ project: worldmem
220
+ mode: offline
221
+ resume: null
222
+ load: null
223
+ auto_resume: true
224
+ resume_ckpt_path: null
225
+ name: train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
226
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
227
+ customized_load: true
228
+ seperate_load: true
229
+ diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
230
+ vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
231
+ only_tune_memory: false
outputs/2026-06-04/06-22-28/.hydra/hydra.yaml ADDED
@@ -0,0 +1,230 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
4
+ sweep:
5
+ dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
6
+ subdir: ${hydra.job.num}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ formatters:
71
+ simple:
72
+ format: '[%(asctime)s][HYDRA] %(message)s'
73
+ handlers:
74
+ console:
75
+ class: logging.StreamHandler
76
+ formatter: simple
77
+ stream: ext://sys.stdout
78
+ root:
79
+ level: INFO
80
+ handlers:
81
+ - console
82
+ loggers:
83
+ logging_example:
84
+ level: DEBUG
85
+ disable_existing_loggers: false
86
+ job_logging:
87
+ version: 1
88
+ formatters:
89
+ simple:
90
+ format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
91
+ handlers:
92
+ console:
93
+ class: logging.StreamHandler
94
+ formatter: simple
95
+ stream: ext://sys.stdout
96
+ file:
97
+ class: logging.FileHandler
98
+ formatter: simple
99
+ filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
100
+ root:
101
+ level: INFO
102
+ handlers:
103
+ - console
104
+ - file
105
+ disable_existing_loggers: false
106
+ env: {}
107
+ mode: RUN
108
+ searchpath: []
109
+ callbacks: {}
110
+ output_subdir: .hydra
111
+ overrides:
112
+ hydra:
113
+ - hydra.mode=RUN
114
+ task:
115
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
116
+ - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
117
+ - auto_resume=true
118
+ - experiment.tasks=[training]
119
+ - algorithm=dememwm_memory_dit
120
+ - +customized_load=true
121
+ - +seperate_load=true
122
+ - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
123
+ - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
124
+ - +only_tune_memory=false
125
+ - dataset=video_minecraft_latent
126
+ - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
127
+ - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
128
+ - dataset.n_frames=1000
129
+ - +dataset.n_frames_valid=500
130
+ - +dataset.customized_validation=true
131
+ - +dataset.wo_updown=false
132
+ - +dataset.angle_range=180
133
+ - +dataset.pos_range=8
134
+ - ++algorithm.n_tokens=4
135
+ - algorithm.x_shape=[16,18,32]
136
+ - ++algorithm.context_frames=100
137
+ - ++algorithm.log_video=true
138
+ - ++algorithm.diffusion.sampling_timesteps=20
139
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
140
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
141
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
142
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
143
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
144
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
145
+ - ++algorithm.dememwm.anchor.enabled=true
146
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
147
+ - ++algorithm.dememwm.anchor.diverse_selection=true
148
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
149
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
150
+ - ++algorithm.dememwm.dynamic.enabled=true
151
+ - ++algorithm.dememwm.dynamic.recent_frames=4
152
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
153
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
154
+ - ++algorithm.dememwm.revisit.enabled=true
155
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
156
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
157
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
158
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
159
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
160
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
161
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
162
+ - ++algorithm.dememwm.revisit.max_frames=2
163
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
164
+ - ++algorithm.dememwm.cache.enabled=true
165
+ - ++algorithm.dememwm.cache.device=cpu
166
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
167
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
168
+ - ++algorithm.dememwm.cache.eviction_policy=none
169
+ - ++algorithm.dememwm.cache.no_evict=true
170
+ - ++algorithm.dememwm.cache.clear_between_videos=true
171
+ - ++algorithm.dememwm.cache.max_records=null
172
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
173
+ - ++algorithm.dememwm.curriculum.enabled=true
174
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
175
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
176
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
177
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
178
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
179
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
180
+ - experiment.training.batch_size=16
181
+ - experiment.training.optim.accumulate_grad_batches=1
182
+ - experiment.validation.batch_size=1
183
+ - experiment.validation.limit_batch=1
184
+ - experiment.training.checkpointing.every_n_train_steps=5000
185
+ - experiment.validation.val_every_n_step=5000
186
+ - experiment.training.max_steps=350000
187
+ job:
188
+ name: main
189
+ chdir: null
190
+ 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
191
+ id: ???
192
+ num: ???
193
+ config_name: training
194
+ env_set: {}
195
+ env_copy: []
196
+ config:
197
+ override_dirname:
198
+ kv_sep: '='
199
+ item_sep: ','
200
+ exclude_keys: []
201
+ runtime:
202
+ version: 1.3.2
203
+ version_base: '1.3'
204
+ cwd: /proj/cvl/users/x_fahkh2/DeMemWM
205
+ config_sources:
206
+ - path: hydra.conf
207
+ schema: pkg
208
+ provider: hydra
209
+ - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations
210
+ schema: file
211
+ provider: main
212
+ - path: ''
213
+ schema: structured
214
+ provider: schema
215
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-04/06-22-28
216
+ choices:
217
+ cluster: null
218
+ algorithm: dememwm_memory_dit
219
+ dataset: video_minecraft_latent
220
+ experiment: exp_video
221
+ hydra/env: default
222
+ hydra/callbacks: null
223
+ hydra/job_logging: default
224
+ hydra/hydra_logging: default
225
+ hydra/hydra_help: default
226
+ hydra/help: default
227
+ hydra/sweeper: basic
228
+ hydra/launcher: basic
229
+ hydra/output: default
230
+ verbose: false
outputs/2026-06-04/06-22-28/.hydra/overrides.yaml ADDED
@@ -0,0 +1,72 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ - +name=train_dememwm_separate_memory_berzelius_8a100_bs16_global128_350k
2
+ - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_separate_memory_berzelius_8a100_bs16_global128_350k/
3
+ - auto_resume=true
4
+ - experiment.tasks=[training]
5
+ - algorithm=dememwm_memory_dit
6
+ - +customized_load=true
7
+ - +seperate_load=true
8
+ - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
9
+ - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
10
+ - +only_tune_memory=false
11
+ - dataset=video_minecraft_latent
12
+ - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
13
+ - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
14
+ - dataset.n_frames=1000
15
+ - +dataset.n_frames_valid=500
16
+ - +dataset.customized_validation=true
17
+ - +dataset.wo_updown=false
18
+ - +dataset.angle_range=180
19
+ - +dataset.pos_range=8
20
+ - ++algorithm.n_tokens=4
21
+ - algorithm.x_shape=[16,18,32]
22
+ - ++algorithm.context_frames=100
23
+ - ++algorithm.log_video=true
24
+ - ++algorithm.diffusion.sampling_timesteps=20
25
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
26
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
27
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
28
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
29
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
30
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
31
+ - ++algorithm.dememwm.anchor.enabled=true
32
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
33
+ - ++algorithm.dememwm.anchor.diverse_selection=true
34
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=2
35
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
36
+ - ++algorithm.dememwm.dynamic.enabled=true
37
+ - ++algorithm.dememwm.dynamic.recent_frames=4
38
+ - ++algorithm.dememwm.dynamic.conv_stride_t=2
39
+ - ++algorithm.dememwm.dynamic.compress.downsample_ratio=2
40
+ - ++algorithm.dememwm.revisit.enabled=true
41
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
42
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
43
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
44
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
45
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
46
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
47
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
48
+ - ++algorithm.dememwm.revisit.max_frames=2
49
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=2
50
+ - ++algorithm.dememwm.cache.enabled=true
51
+ - ++algorithm.dememwm.cache.device=cpu
52
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
53
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
54
+ - ++algorithm.dememwm.cache.eviction_policy=none
55
+ - ++algorithm.dememwm.cache.no_evict=true
56
+ - ++algorithm.dememwm.cache.clear_between_videos=true
57
+ - ++algorithm.dememwm.cache.max_records=null
58
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
59
+ - ++algorithm.dememwm.curriculum.enabled=true
60
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
61
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
62
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
63
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
64
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
65
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
66
+ - experiment.training.batch_size=16
67
+ - experiment.training.optim.accumulate_grad_batches=1
68
+ - experiment.validation.batch_size=1
69
+ - experiment.validation.limit_batch=1
70
+ - experiment.training.checkpointing.every_n_train_steps=5000
71
+ - experiment.validation.val_every_n_step=5000
72
+ - experiment.training.max_steps=350000
outputs/2026-06-04/06-22-28/main.log ADDED
File without changes
outputs/2026-06-04/06-24-27/.hydra/config.yaml ADDED
@@ -0,0 +1,232 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment:
2
+ debug: ${debug}
3
+ tasks:
4
+ - training
5
+ num_nodes: 1
6
+ training:
7
+ precision: 16-mixed
8
+ compile: false
9
+ lr: 2.0e-05
10
+ batch_size: 16
11
+ max_epochs: -1
12
+ max_steps: 350000
13
+ max_time: null
14
+ data:
15
+ num_workers: 4
16
+ shuffle: true
17
+ optim:
18
+ accumulate_grad_batches: 1
19
+ gradient_clip_val: 1.0
20
+ checkpointing:
21
+ every_n_train_steps: 5000
22
+ every_n_epochs: null
23
+ train_time_interval: null
24
+ save_last: true
25
+ enable_version_counter: false
26
+ validation:
27
+ precision: 16-mixed
28
+ compile: false
29
+ batch_size: 1
30
+ val_every_n_step: 5000
31
+ val_every_n_epoch: null
32
+ limit_batch: 1
33
+ inference_mode: true
34
+ data:
35
+ num_workers: 4
36
+ shuffle: false
37
+ test:
38
+ precision: 16-mixed
39
+ compile: false
40
+ batch_size: 1
41
+ limit_batch: 1
42
+ data:
43
+ num_workers: 4
44
+ shuffle: false
45
+ logging:
46
+ metrics: null
47
+ dataset:
48
+ debug: ${debug}
49
+ metadata: data/${dataset.name}/metadata.json
50
+ data_mean: 0.5
51
+ data_std: 0.5
52
+ save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
53
+ n_frames: 1000
54
+ context_length: 1
55
+ resolution: 128
56
+ observation_shape:
57
+ - 3
58
+ - ${dataset.resolution}
59
+ - ${dataset.resolution}
60
+ external_cond_dim: 0
61
+ validation_multiplier: 1
62
+ frame_skip: 1
63
+ action_cond_dim: 25
64
+ _name: video_minecraft_latent
65
+ precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
66
+ n_frames_valid: 500
67
+ customized_validation: true
68
+ memory_condition_length: 0
69
+ wo_updown: false
70
+ angle_range: 180
71
+ pos_range: 8
72
+ algorithm:
73
+ debug: ${debug}
74
+ lr: ${experiment.training.lr}
75
+ x_shape:
76
+ - 16
77
+ - 18
78
+ - 32
79
+ frame_stack: 1
80
+ frame_skip: ${dataset.frame_skip}
81
+ data_mean: ${dataset.data_mean}
82
+ data_std: ${dataset.data_std}
83
+ external_cond_dim: 0
84
+ context_frames: 100
85
+ weight_decay: 0.002
86
+ warmup_steps: 1000
87
+ optimizer_beta:
88
+ - 0.9
89
+ - 0.99
90
+ uncertainty_scale: 1
91
+ guidance_scale: 0.0
92
+ chunk_size: 1
93
+ scheduling_matrix: autoregressive
94
+ noise_level: random_all
95
+ causal: true
96
+ diffusion:
97
+ objective: pred_v
98
+ beta_schedule: sigmoid
99
+ schedule_fn_kwargs: {}
100
+ clip_noise: 20.0
101
+ use_snr: false
102
+ use_cum_snr: false
103
+ use_fused_snr: true
104
+ snr_clip: 5.0
105
+ cum_snr_decay: 0.96
106
+ timesteps: 1000
107
+ sampling_timesteps: 20
108
+ ddim_sampling_eta: 0.0
109
+ stabilization_level: 15
110
+ architecture:
111
+ network_size: 64
112
+ attn_heads: 4
113
+ attn_dim_head: 64
114
+ dim_mults:
115
+ - 1
116
+ - 2
117
+ - 4
118
+ - 8
119
+ resolution: ${dataset.resolution}
120
+ attn_resolutions:
121
+ - 16
122
+ - 32
123
+ - 64
124
+ - 128
125
+ use_init_temporal_attn: true
126
+ use_linear_attn: true
127
+ time_emb_type: rotary
128
+ n_frames: ${dataset.n_frames}
129
+ metadata: ${dataset.metadata}
130
+ action_cond_dim: 25
131
+ use_plucker: true
132
+ _name: dememwm_memory_dit
133
+ memory_token_cross_attention: true
134
+ memory_cross_attn_layers: null
135
+ memory_condition_length: 0
136
+ pose_cond_dim: 5
137
+ log_video: true
138
+ dememwm:
139
+ enabled: true
140
+ training_stage: stage_1
141
+ curriculum:
142
+ enabled: true
143
+ full_stage_start_step: 20000
144
+ freeze_vae: true
145
+ dit_freeze:
146
+ enabled: true
147
+ lr:
148
+ dememwm_modules: 4.0e-05
149
+ memory_adapters: 4.0e-05
150
+ full_dit: 1.0e-05
151
+ token_patch_size: 2
152
+ anchor:
153
+ enabled: true
154
+ anchor_indices:
155
+ - 0
156
+ - 1
157
+ - 2
158
+ - 3
159
+ allow_generated_as_anchor: false
160
+ diverse_selection: true
161
+ compress:
162
+ downsample_ratio: 3
163
+ dynamic:
164
+ enabled: true
165
+ exclude_latest_local_frames: 4
166
+ recent_frames: 4
167
+ conv_kernel_t: 3
168
+ conv_stride_t: 2
169
+ revisit:
170
+ enabled: true
171
+ deterministic_pose_retrieval: true
172
+ fov_overlap_threshold: 0.6
173
+ high_quality_fov_threshold: 0.7
174
+ plucker_weight: 0.1
175
+ max_frames: 2
176
+ fov_half_h: 52.5
177
+ fov_half_v: 37.5
178
+ fov_radius: 30.0
179
+ fov_yaw_samples: 25
180
+ fov_pitch_samples: 20
181
+ fov_depth_samples: 20
182
+ pose_preselect_topk: 64
183
+ plucker_grid_h: 4
184
+ plucker_grid_w: 4
185
+ plucker_focal_length: 0.35
186
+ compress:
187
+ downsample_ratio: 3
188
+ eval_ablation:
189
+ enabled: false
190
+ branch: A_plus_D_plus_R_normal
191
+ generated_history_proxy:
192
+ enabled: true
193
+ start_step: 40000
194
+ ramp_steps: 40000
195
+ max_prob: 0.25
196
+ noise_std: 0.25
197
+ dropout_prob: 0.0
198
+ injection:
199
+ dit_hidden_size: 1024
200
+ anchor_gate: 1.0
201
+ dynamic_gate: 1.0
202
+ revisit_gate: 1.0
203
+ cache:
204
+ enabled: true
205
+ device: cpu
206
+ keep_raw_latents: all
207
+ keep_compressed_records: true
208
+ keep_prefix_anchors: true
209
+ eviction_policy: none
210
+ no_evict: true
211
+ clear_between_videos: true
212
+ max_records: null
213
+ on_capacity_exceeded: warn
214
+ checkpoint:
215
+ strict_dememwm_eval_load: true
216
+ n_tokens: 4
217
+ debug: false
218
+ wandb:
219
+ entity: turlin
220
+ project: worldmem
221
+ mode: offline
222
+ resume: null
223
+ load: null
224
+ auto_resume: true
225
+ resume_ckpt_path: null
226
+ name: train_dememwm_full_berzelius_8a100_bs16_global128_350k
227
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/
228
+ customized_load: true
229
+ seperate_load: true
230
+ diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
231
+ vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
232
+ only_tune_memory: false
outputs/2026-06-04/06-24-27/.hydra/hydra.yaml ADDED
@@ -0,0 +1,230 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
4
+ sweep:
5
+ dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
6
+ subdir: ${hydra.job.num}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ formatters:
71
+ simple:
72
+ format: '[%(asctime)s][HYDRA] %(message)s'
73
+ handlers:
74
+ console:
75
+ class: logging.StreamHandler
76
+ formatter: simple
77
+ stream: ext://sys.stdout
78
+ root:
79
+ level: INFO
80
+ handlers:
81
+ - console
82
+ loggers:
83
+ logging_example:
84
+ level: DEBUG
85
+ disable_existing_loggers: false
86
+ job_logging:
87
+ version: 1
88
+ formatters:
89
+ simple:
90
+ format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
91
+ handlers:
92
+ console:
93
+ class: logging.StreamHandler
94
+ formatter: simple
95
+ stream: ext://sys.stdout
96
+ file:
97
+ class: logging.FileHandler
98
+ formatter: simple
99
+ filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
100
+ root:
101
+ level: INFO
102
+ handlers:
103
+ - console
104
+ - file
105
+ disable_existing_loggers: false
106
+ env: {}
107
+ mode: RUN
108
+ searchpath: []
109
+ callbacks: {}
110
+ output_subdir: .hydra
111
+ overrides:
112
+ hydra:
113
+ - hydra.mode=RUN
114
+ task:
115
+ - +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k
116
+ - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/
117
+ - auto_resume=true
118
+ - experiment.tasks=[training]
119
+ - algorithm=dememwm_memory_dit
120
+ - +customized_load=true
121
+ - +seperate_load=true
122
+ - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
123
+ - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
124
+ - +only_tune_memory=false
125
+ - dataset=video_minecraft_latent
126
+ - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
127
+ - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
128
+ - dataset.n_frames=1000
129
+ - +dataset.n_frames_valid=500
130
+ - +dataset.customized_validation=true
131
+ - +dataset.memory_condition_length=0
132
+ - +dataset.wo_updown=false
133
+ - +dataset.angle_range=180
134
+ - +dataset.pos_range=8
135
+ - ++algorithm.n_tokens=4
136
+ - algorithm.x_shape=[16,18,32]
137
+ - ++algorithm.context_frames=100
138
+ - ++algorithm.log_video=true
139
+ - ++algorithm.diffusion.sampling_timesteps=20
140
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
141
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
142
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
143
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
144
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
145
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
146
+ - ++algorithm.dememwm.anchor.enabled=true
147
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
148
+ - ++algorithm.dememwm.anchor.diverse_selection=true
149
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=3
150
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
151
+ - ++algorithm.dememwm.dynamic.enabled=true
152
+ - ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4
153
+ - ++algorithm.dememwm.dynamic.recent_frames=4
154
+ - ++algorithm.dememwm.revisit.enabled=true
155
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
156
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
157
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
158
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
159
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
160
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
161
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
162
+ - ++algorithm.dememwm.revisit.max_frames=2
163
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=3
164
+ - ++algorithm.dememwm.cache.enabled=true
165
+ - ++algorithm.dememwm.cache.device=cpu
166
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
167
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
168
+ - ++algorithm.dememwm.cache.eviction_policy=none
169
+ - ++algorithm.dememwm.cache.no_evict=true
170
+ - ++algorithm.dememwm.cache.clear_between_videos=true
171
+ - ++algorithm.dememwm.cache.max_records=null
172
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
173
+ - ++algorithm.dememwm.curriculum.enabled=true
174
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
175
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
176
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
177
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
178
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
179
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
180
+ - experiment.training.batch_size=16
181
+ - experiment.training.optim.accumulate_grad_batches=1
182
+ - experiment.validation.batch_size=1
183
+ - experiment.validation.limit_batch=1
184
+ - experiment.training.checkpointing.every_n_train_steps=5000
185
+ - experiment.validation.val_every_n_step=5000
186
+ - experiment.training.max_steps=350000
187
+ job:
188
+ name: main
189
+ chdir: null
190
+ 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
191
+ id: ???
192
+ num: ???
193
+ config_name: training
194
+ env_set: {}
195
+ env_copy: []
196
+ config:
197
+ override_dirname:
198
+ kv_sep: '='
199
+ item_sep: ','
200
+ exclude_keys: []
201
+ runtime:
202
+ version: 1.3.2
203
+ version_base: '1.3'
204
+ cwd: /proj/cvl/users/x_fahkh2/DeMemWM
205
+ config_sources:
206
+ - path: hydra.conf
207
+ schema: pkg
208
+ provider: hydra
209
+ - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations
210
+ schema: file
211
+ provider: main
212
+ - path: ''
213
+ schema: structured
214
+ provider: schema
215
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-04/06-24-27
216
+ choices:
217
+ cluster: null
218
+ algorithm: dememwm_memory_dit
219
+ dataset: video_minecraft_latent
220
+ experiment: exp_video
221
+ hydra/env: default
222
+ hydra/callbacks: null
223
+ hydra/job_logging: default
224
+ hydra/hydra_logging: default
225
+ hydra/hydra_help: default
226
+ hydra/help: default
227
+ hydra/sweeper: basic
228
+ hydra/launcher: basic
229
+ hydra/output: default
230
+ verbose: false
outputs/2026-06-04/06-24-27/.hydra/overrides.yaml ADDED
@@ -0,0 +1,72 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ - +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k
2
+ - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/
3
+ - auto_resume=true
4
+ - experiment.tasks=[training]
5
+ - algorithm=dememwm_memory_dit
6
+ - +customized_load=true
7
+ - +seperate_load=true
8
+ - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
9
+ - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
10
+ - +only_tune_memory=false
11
+ - dataset=video_minecraft_latent
12
+ - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
13
+ - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
14
+ - dataset.n_frames=1000
15
+ - +dataset.n_frames_valid=500
16
+ - +dataset.customized_validation=true
17
+ - +dataset.memory_condition_length=0
18
+ - +dataset.wo_updown=false
19
+ - +dataset.angle_range=180
20
+ - +dataset.pos_range=8
21
+ - ++algorithm.n_tokens=4
22
+ - algorithm.x_shape=[16,18,32]
23
+ - ++algorithm.context_frames=100
24
+ - ++algorithm.log_video=true
25
+ - ++algorithm.diffusion.sampling_timesteps=20
26
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
27
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
28
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
29
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
30
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
31
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
32
+ - ++algorithm.dememwm.anchor.enabled=true
33
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
34
+ - ++algorithm.dememwm.anchor.diverse_selection=true
35
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=3
36
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
37
+ - ++algorithm.dememwm.dynamic.enabled=true
38
+ - ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4
39
+ - ++algorithm.dememwm.dynamic.recent_frames=4
40
+ - ++algorithm.dememwm.revisit.enabled=true
41
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
42
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
43
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
44
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
45
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
46
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
47
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
48
+ - ++algorithm.dememwm.revisit.max_frames=2
49
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=3
50
+ - ++algorithm.dememwm.cache.enabled=true
51
+ - ++algorithm.dememwm.cache.device=cpu
52
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
53
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
54
+ - ++algorithm.dememwm.cache.eviction_policy=none
55
+ - ++algorithm.dememwm.cache.no_evict=true
56
+ - ++algorithm.dememwm.cache.clear_between_videos=true
57
+ - ++algorithm.dememwm.cache.max_records=null
58
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
59
+ - ++algorithm.dememwm.curriculum.enabled=true
60
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
61
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
62
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
63
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
64
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
65
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
66
+ - experiment.training.batch_size=16
67
+ - experiment.training.optim.accumulate_grad_batches=1
68
+ - experiment.validation.batch_size=1
69
+ - experiment.validation.limit_batch=1
70
+ - experiment.training.checkpointing.every_n_train_steps=5000
71
+ - experiment.validation.val_every_n_step=5000
72
+ - experiment.training.max_steps=350000
outputs/2026-06-04/06-24-27/main.log ADDED
@@ -0,0 +1,77 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [2026-06-04 06:24:58,261][pytorch_lightning.utilities.rank_zero][INFO] - Using 16bit Automatic Mixed Precision (AMP)
2
+ [2026-06-04 06:24:58,374][pytorch_lightning.utilities.rank_zero][INFO] - GPU available: True (cuda), used: True
3
+ [2026-06-04 06:24:58,374][pytorch_lightning.utilities.rank_zero][INFO] - TPU available: False, using: 0 TPU cores
4
+ [2026-06-04 06:24:58,374][pytorch_lightning.utilities.rank_zero][INFO] - IPU available: False, using: 0 IPUs
5
+ [2026-06-04 06:24:58,374][pytorch_lightning.utilities.rank_zero][INFO] - HPU available: False, using: 0 HPUs
6
+ [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.
7
+ [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=''
8
+ [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=''
9
+ [2026-06-04 06:25:01,130][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 0, MEMBER: 1/8
10
+ [2026-06-04 06:25:01,209][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 7, MEMBER: 8/8
11
+ [2026-06-04 06:25:01,432][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 2, MEMBER: 3/8
12
+ [2026-06-04 06:25:01,930][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 6, MEMBER: 7/8
13
+ [2026-06-04 06:25:02,129][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 4, MEMBER: 5/8
14
+ [2026-06-04 06:25:02,209][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 3, MEMBER: 4/8
15
+ [2026-06-04 06:25:03,309][lightning.fabric.utilities.distributed][INFO] - Initializing distributed: GLOBAL_RANK: 5, MEMBER: 6/8
16
+ [2026-06-04 06:25:07,333][pytorch_lightning.utilities.rank_zero][INFO] - ----------------------------------------------------------------------------------------------------
17
+ distributed_backend=nccl
18
+ All distributed processes registered. Starting with 8 processes
19
+ ----------------------------------------------------------------------------------------------------
20
+
21
+ [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]
22
+ [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]
23
+ [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]
24
+ [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]
25
+ [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]
26
+ [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]
27
+ [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]
28
+ [2026-06-04 06:25:21,907][lightning.pytorch.callbacks.model_summary][INFO] -
29
+ | Name | Type | Params
30
+ -------------------------------------------------------------------------------------
31
+ 0 | diffusion_model | Diffusion | 910 M
32
+ 1 | dememwm_dynamic_compressor | CausalConv3DDynamicCompressor | 199 K
33
+ 2 | dememwm_anchor_proj | SpatialConv2DMemoryProjector | 591 K
34
+ 3 | dememwm_revisit_proj | SpatialConv2DMemoryProjector | 591 K
35
+ 4 | dememwm_revisit_gate | RevisitRawGate | 4
36
+ 5 | validation_lpips_model | LearnedPerceptualImagePatchSimilarity | 2.5 M
37
+ 6 | vae | AutoencoderKL | 229 M
38
+ -------------------------------------------------------------------------------------
39
+ 911 M Trainable params
40
+ 231 M Non-trainable params
41
+ 1.1 B Total params
42
+ 4,573.841 Total estimated model params size (MB)
43
+ [2026-06-04 06:25:22,815][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
44
+ [2026-06-04 06:25:22,815][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
45
+ [2026-06-04 06:25:22,815][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
46
+ [2026-06-04 06:25:22,815][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
47
+ [2026-06-04 06:25:22,815][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
48
+ [2026-06-04 06:25:22,815][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
49
+ [2026-06-04 06:25:22,815][lightning.pytorch.trainer.connectors.signal_connector][INFO] - SLURM auto-requeueing enabled. Setting signal handlers.
50
+ [2026-06-07 06:24:48,015][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15
51
+ [2026-06-07 06:24:48,015][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15
52
+ [2026-06-07 06:24:48,016][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
53
+ [2026-06-07 06:24:48,015][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15
54
+ [2026-06-07 06:24:48,016][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
55
+ [2026-06-07 06:24:48,015][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15
56
+ [2026-06-07 06:24:48,016][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
57
+ [2026-06-07 06:24:48,016][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
58
+ [2026-06-07 06:24:48,015][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15
59
+ [2026-06-07 06:24:48,016][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
60
+ [2026-06-07 06:24:48,016][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15
61
+ [2026-06-07 06:24:48,015][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15
62
+ [2026-06-07 06:24:48,016][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
63
+ [2026-06-07 06:24:48,016][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
64
+ [2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 2] Received SIGTERM: 15
65
+ [2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 0] Received SIGTERM: 15
66
+ [2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
67
+ [2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 3] Received SIGTERM: 15
68
+ [2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
69
+ [2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 4] Received SIGTERM: 15
70
+ [2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
71
+ [2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
72
+ [2026-06-07 06:24:48,020][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 5] Received SIGTERM: 15
73
+ [2026-06-07 06:24:48,021][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
74
+ [2026-06-07 06:24:48,021][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 6] Received SIGTERM: 15
75
+ [2026-06-07 06:24:48,021][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
76
+ [2026-06-07 06:24:48,021][lightning.pytorch.trainer.connectors.signal_connector][INFO] - [rank: 7] Received SIGTERM: 15
77
+ [2026-06-07 06:24:48,021][lightning.pytorch.trainer.connectors.signal_connector][INFO] - Bypassing SIGTERM: 15
outputs/2026-06-04/06-24-28/.hydra/config.yaml ADDED
@@ -0,0 +1,232 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment:
2
+ debug: ${debug}
3
+ tasks:
4
+ - training
5
+ num_nodes: 1
6
+ training:
7
+ precision: 16-mixed
8
+ compile: false
9
+ lr: 2.0e-05
10
+ batch_size: 16
11
+ max_epochs: -1
12
+ max_steps: 350000
13
+ max_time: null
14
+ data:
15
+ num_workers: 4
16
+ shuffle: true
17
+ optim:
18
+ accumulate_grad_batches: 1
19
+ gradient_clip_val: 1.0
20
+ checkpointing:
21
+ every_n_train_steps: 5000
22
+ every_n_epochs: null
23
+ train_time_interval: null
24
+ save_last: true
25
+ enable_version_counter: false
26
+ validation:
27
+ precision: 16-mixed
28
+ compile: false
29
+ batch_size: 1
30
+ val_every_n_step: 5000
31
+ val_every_n_epoch: null
32
+ limit_batch: 1
33
+ inference_mode: true
34
+ data:
35
+ num_workers: 4
36
+ shuffle: false
37
+ test:
38
+ precision: 16-mixed
39
+ compile: false
40
+ batch_size: 1
41
+ limit_batch: 1
42
+ data:
43
+ num_workers: 4
44
+ shuffle: false
45
+ logging:
46
+ metrics: null
47
+ dataset:
48
+ debug: ${debug}
49
+ metadata: data/${dataset.name}/metadata.json
50
+ data_mean: 0.5
51
+ data_std: 0.5
52
+ save_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
53
+ n_frames: 1000
54
+ context_length: 1
55
+ resolution: 128
56
+ observation_shape:
57
+ - 3
58
+ - ${dataset.resolution}
59
+ - ${dataset.resolution}
60
+ external_cond_dim: 0
61
+ validation_multiplier: 1
62
+ frame_skip: 1
63
+ action_cond_dim: 25
64
+ _name: video_minecraft_latent
65
+ precomputed_feature_dir: /proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
66
+ n_frames_valid: 500
67
+ customized_validation: true
68
+ memory_condition_length: 0
69
+ wo_updown: false
70
+ angle_range: 180
71
+ pos_range: 8
72
+ algorithm:
73
+ debug: ${debug}
74
+ lr: ${experiment.training.lr}
75
+ x_shape:
76
+ - 16
77
+ - 18
78
+ - 32
79
+ frame_stack: 1
80
+ frame_skip: ${dataset.frame_skip}
81
+ data_mean: ${dataset.data_mean}
82
+ data_std: ${dataset.data_std}
83
+ external_cond_dim: 0
84
+ context_frames: 100
85
+ weight_decay: 0.002
86
+ warmup_steps: 1000
87
+ optimizer_beta:
88
+ - 0.9
89
+ - 0.99
90
+ uncertainty_scale: 1
91
+ guidance_scale: 0.0
92
+ chunk_size: 1
93
+ scheduling_matrix: autoregressive
94
+ noise_level: random_all
95
+ causal: true
96
+ diffusion:
97
+ objective: pred_v
98
+ beta_schedule: sigmoid
99
+ schedule_fn_kwargs: {}
100
+ clip_noise: 20.0
101
+ use_snr: false
102
+ use_cum_snr: false
103
+ use_fused_snr: true
104
+ snr_clip: 5.0
105
+ cum_snr_decay: 0.96
106
+ timesteps: 1000
107
+ sampling_timesteps: 20
108
+ ddim_sampling_eta: 0.0
109
+ stabilization_level: 15
110
+ architecture:
111
+ network_size: 64
112
+ attn_heads: 4
113
+ attn_dim_head: 64
114
+ dim_mults:
115
+ - 1
116
+ - 2
117
+ - 4
118
+ - 8
119
+ resolution: ${dataset.resolution}
120
+ attn_resolutions:
121
+ - 16
122
+ - 32
123
+ - 64
124
+ - 128
125
+ use_init_temporal_attn: true
126
+ use_linear_attn: true
127
+ time_emb_type: rotary
128
+ n_frames: ${dataset.n_frames}
129
+ metadata: ${dataset.metadata}
130
+ action_cond_dim: 25
131
+ use_plucker: true
132
+ _name: dememwm_memory_dit
133
+ memory_token_cross_attention: true
134
+ memory_cross_attn_layers: null
135
+ memory_condition_length: 0
136
+ pose_cond_dim: 5
137
+ log_video: true
138
+ dememwm:
139
+ enabled: true
140
+ training_stage: stage_1
141
+ curriculum:
142
+ enabled: true
143
+ full_stage_start_step: 20000
144
+ freeze_vae: true
145
+ dit_freeze:
146
+ enabled: true
147
+ lr:
148
+ dememwm_modules: 4.0e-05
149
+ memory_adapters: 4.0e-05
150
+ full_dit: 1.0e-05
151
+ token_patch_size: 2
152
+ anchor:
153
+ enabled: true
154
+ anchor_indices:
155
+ - 0
156
+ - 1
157
+ - 2
158
+ - 3
159
+ allow_generated_as_anchor: false
160
+ diverse_selection: true
161
+ compress:
162
+ downsample_ratio: 3
163
+ dynamic:
164
+ enabled: true
165
+ exclude_latest_local_frames: 4
166
+ recent_frames: 4
167
+ conv_kernel_t: 3
168
+ conv_stride_t: 2
169
+ revisit:
170
+ enabled: true
171
+ deterministic_pose_retrieval: true
172
+ fov_overlap_threshold: 0.6
173
+ high_quality_fov_threshold: 0.7
174
+ plucker_weight: 0.1
175
+ max_frames: 2
176
+ fov_half_h: 52.5
177
+ fov_half_v: 37.5
178
+ fov_radius: 30.0
179
+ fov_yaw_samples: 25
180
+ fov_pitch_samples: 20
181
+ fov_depth_samples: 20
182
+ pose_preselect_topk: 64
183
+ plucker_grid_h: 4
184
+ plucker_grid_w: 4
185
+ plucker_focal_length: 0.35
186
+ compress:
187
+ downsample_ratio: 3
188
+ eval_ablation:
189
+ enabled: false
190
+ branch: A_plus_D_plus_R_normal
191
+ generated_history_proxy:
192
+ enabled: true
193
+ start_step: 40000
194
+ ramp_steps: 40000
195
+ max_prob: 0.25
196
+ noise_std: 0.25
197
+ dropout_prob: 0.0
198
+ injection:
199
+ dit_hidden_size: 1024
200
+ anchor_gate: 1.0
201
+ dynamic_gate: 1.0
202
+ revisit_gate: 1.0
203
+ cache:
204
+ enabled: true
205
+ device: cpu
206
+ keep_raw_latents: all
207
+ keep_compressed_records: true
208
+ keep_prefix_anchors: true
209
+ eviction_policy: none
210
+ no_evict: true
211
+ clear_between_videos: true
212
+ max_records: null
213
+ on_capacity_exceeded: warn
214
+ checkpoint:
215
+ strict_dememwm_eval_load: true
216
+ n_tokens: 4
217
+ debug: false
218
+ wandb:
219
+ entity: turlin
220
+ project: worldmem
221
+ mode: offline
222
+ resume: null
223
+ load: null
224
+ auto_resume: true
225
+ resume_ckpt_path: null
226
+ name: train_dememwm_full_berzelius_8a100_bs16_global128_350k
227
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/
228
+ customized_load: true
229
+ seperate_load: true
230
+ diffusion_model_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
231
+ vae_path: /proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
232
+ only_tune_memory: false
outputs/2026-06-04/06-24-28/.hydra/hydra.yaml ADDED
@@ -0,0 +1,230 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ hydra:
2
+ run:
3
+ dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
4
+ sweep:
5
+ dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
6
+ subdir: ${hydra.job.num}
7
+ launcher:
8
+ _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
9
+ sweeper:
10
+ _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
11
+ max_batch_size: null
12
+ params: null
13
+ help:
14
+ app_name: ${hydra.job.name}
15
+ header: '${hydra.help.app_name} is powered by Hydra.
16
+
17
+ '
18
+ footer: 'Powered by Hydra (https://hydra.cc)
19
+
20
+ Use --hydra-help to view Hydra specific help
21
+
22
+ '
23
+ template: '${hydra.help.header}
24
+
25
+ == Configuration groups ==
26
+
27
+ Compose your configuration from those groups (group=option)
28
+
29
+
30
+ $APP_CONFIG_GROUPS
31
+
32
+
33
+ == Config ==
34
+
35
+ Override anything in the config (foo.bar=value)
36
+
37
+
38
+ $CONFIG
39
+
40
+
41
+ ${hydra.help.footer}
42
+
43
+ '
44
+ hydra_help:
45
+ template: 'Hydra (${hydra.runtime.version})
46
+
47
+ See https://hydra.cc for more info.
48
+
49
+
50
+ == Flags ==
51
+
52
+ $FLAGS_HELP
53
+
54
+
55
+ == Configuration groups ==
56
+
57
+ Compose your configuration from those groups (For example, append hydra/job_logging=disabled
58
+ to command line)
59
+
60
+
61
+ $HYDRA_CONFIG_GROUPS
62
+
63
+
64
+ Use ''--cfg hydra'' to Show the Hydra config.
65
+
66
+ '
67
+ hydra_help: ???
68
+ hydra_logging:
69
+ version: 1
70
+ formatters:
71
+ simple:
72
+ format: '[%(asctime)s][HYDRA] %(message)s'
73
+ handlers:
74
+ console:
75
+ class: logging.StreamHandler
76
+ formatter: simple
77
+ stream: ext://sys.stdout
78
+ root:
79
+ level: INFO
80
+ handlers:
81
+ - console
82
+ loggers:
83
+ logging_example:
84
+ level: DEBUG
85
+ disable_existing_loggers: false
86
+ job_logging:
87
+ version: 1
88
+ formatters:
89
+ simple:
90
+ format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
91
+ handlers:
92
+ console:
93
+ class: logging.StreamHandler
94
+ formatter: simple
95
+ stream: ext://sys.stdout
96
+ file:
97
+ class: logging.FileHandler
98
+ formatter: simple
99
+ filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
100
+ root:
101
+ level: INFO
102
+ handlers:
103
+ - console
104
+ - file
105
+ disable_existing_loggers: false
106
+ env: {}
107
+ mode: RUN
108
+ searchpath: []
109
+ callbacks: {}
110
+ output_subdir: .hydra
111
+ overrides:
112
+ hydra:
113
+ - hydra.mode=RUN
114
+ task:
115
+ - +name=train_dememwm_full_berzelius_8a100_bs16_global128_350k
116
+ - +output_dir=/proj/cvl/users/x_fahkh2/DeMemWM/results/dememwm_full_berzelius_8a100_bs16_global128_350k/
117
+ - auto_resume=true
118
+ - experiment.tasks=[training]
119
+ - algorithm=dememwm_memory_dit
120
+ - +customized_load=true
121
+ - +seperate_load=true
122
+ - +diffusion_model_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/oasis500m.safetensors
123
+ - +vae_path=/proj/cvl/users/x_fahkh2/WorldMem_Repro/checkpoints/vit-l-20.safetensors
124
+ - +only_tune_memory=false
125
+ - dataset=video_minecraft_latent
126
+ - dataset.save_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft
127
+ - dataset.precomputed_feature_dir=/proj/cvl/users/x_fahkh2/WorldMem_Repro/datasets/minecraft/vae_features
128
+ - dataset.n_frames=1000
129
+ - +dataset.n_frames_valid=500
130
+ - +dataset.customized_validation=true
131
+ - +dataset.memory_condition_length=0
132
+ - +dataset.wo_updown=false
133
+ - +dataset.angle_range=180
134
+ - +dataset.pos_range=8
135
+ - ++algorithm.n_tokens=4
136
+ - algorithm.x_shape=[16,18,32]
137
+ - ++algorithm.context_frames=100
138
+ - ++algorithm.log_video=true
139
+ - ++algorithm.diffusion.sampling_timesteps=20
140
+ - ++algorithm.dememwm.generated_history_proxy.enabled=true
141
+ - ++algorithm.dememwm.generated_history_proxy.start_step=40000
142
+ - ++algorithm.dememwm.generated_history_proxy.ramp_steps=40000
143
+ - ++algorithm.dememwm.generated_history_proxy.max_prob=0.25
144
+ - ++algorithm.dememwm.generated_history_proxy.noise_std=0.25
145
+ - ++algorithm.dememwm.generated_history_proxy.dropout_prob=0.0
146
+ - ++algorithm.dememwm.anchor.enabled=true
147
+ - ++algorithm.dememwm.anchor.anchor_indices=[0,1,2,3]
148
+ - ++algorithm.dememwm.anchor.diverse_selection=true
149
+ - ++algorithm.dememwm.anchor.compress.downsample_ratio=3
150
+ - ++algorithm.dememwm.anchor.allow_generated_as_anchor=false
151
+ - ++algorithm.dememwm.dynamic.enabled=true
152
+ - ++algorithm.dememwm.dynamic.exclude_latest_local_frames=4
153
+ - ++algorithm.dememwm.dynamic.recent_frames=4
154
+ - ++algorithm.dememwm.revisit.enabled=true
155
+ - ++algorithm.dememwm.revisit.deterministic_pose_retrieval=true
156
+ - ++algorithm.dememwm.revisit.fov_overlap_threshold=0.60
157
+ - ++algorithm.dememwm.revisit.pose_preselect_topk=64
158
+ - ++algorithm.dememwm.revisit.fov_yaw_samples=25
159
+ - ++algorithm.dememwm.revisit.fov_pitch_samples=20
160
+ - ++algorithm.dememwm.revisit.fov_depth_samples=20
161
+ - ++algorithm.dememwm.revisit.plucker_weight=0.10
162
+ - ++algorithm.dememwm.revisit.max_frames=2
163
+ - ++algorithm.dememwm.revisit.compress.downsample_ratio=3
164
+ - ++algorithm.dememwm.cache.enabled=true
165
+ - ++algorithm.dememwm.cache.device=cpu
166
+ - ++algorithm.dememwm.cache.keep_raw_latents=all
167
+ - ++algorithm.dememwm.cache.keep_compressed_records=true
168
+ - ++algorithm.dememwm.cache.eviction_policy=none
169
+ - ++algorithm.dememwm.cache.no_evict=true
170
+ - ++algorithm.dememwm.cache.clear_between_videos=true
171
+ - ++algorithm.dememwm.cache.max_records=null
172
+ - ++algorithm.dememwm.cache.on_capacity_exceeded=warn
173
+ - ++algorithm.dememwm.curriculum.enabled=true
174
+ - ++algorithm.dememwm.curriculum.full_stage_start_step=20000
175
+ - ++algorithm.dememwm.curriculum.freeze_vae=true
176
+ - ++algorithm.dememwm.curriculum.dit_freeze.enabled=true
177
+ - ++algorithm.dememwm.curriculum.lr.dememwm_modules=4.0e-5
178
+ - ++algorithm.dememwm.curriculum.lr.memory_adapters=4.0e-5
179
+ - ++algorithm.dememwm.curriculum.lr.full_dit=1.0e-5
180
+ - experiment.training.batch_size=16
181
+ - experiment.training.optim.accumulate_grad_batches=1
182
+ - experiment.validation.batch_size=1
183
+ - experiment.validation.limit_batch=1
184
+ - experiment.training.checkpointing.every_n_train_steps=5000
185
+ - experiment.validation.val_every_n_step=5000
186
+ - experiment.training.max_steps=350000
187
+ job:
188
+ name: main
189
+ chdir: null
190
+ 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
191
+ id: ???
192
+ num: ???
193
+ config_name: training
194
+ env_set: {}
195
+ env_copy: []
196
+ config:
197
+ override_dirname:
198
+ kv_sep: '='
199
+ item_sep: ','
200
+ exclude_keys: []
201
+ runtime:
202
+ version: 1.3.2
203
+ version_base: '1.3'
204
+ cwd: /proj/cvl/users/x_fahkh2/DeMemWM
205
+ config_sources:
206
+ - path: hydra.conf
207
+ schema: pkg
208
+ provider: hydra
209
+ - path: /proj/cvl/users/x_fahkh2/DeMemWM/configurations
210
+ schema: file
211
+ provider: main
212
+ - path: ''
213
+ schema: structured
214
+ provider: schema
215
+ output_dir: /proj/cvl/users/x_fahkh2/DeMemWM/outputs/2026-06-04/06-24-28
216
+ choices:
217
+ cluster: null
218
+ algorithm: dememwm_memory_dit
219
+ dataset: video_minecraft_latent
220
+ experiment: exp_video
221
+ hydra/env: default
222
+ hydra/callbacks: null
223
+ hydra/job_logging: default
224
+ hydra/hydra_logging: default
225
+ hydra/hydra_help: default
226
+ hydra/help: default
227
+ hydra/sweeper: basic
228
+ hydra/launcher: basic
229
+ hydra/output: default
230
+ verbose: false