Publish atlantis qwenpi_v3 H8 training artifacts

zero-latency/atlantis/vla/starvla-qwenpi_v3-h1/atlantis-qwenpi_v3-hist8-zero-s42-v1/README.md

@@ -0,0 +1,7 @@
+# Atlantis qwenpi_v3 zero-latency H8 VLA
+
+Current RGB and eight binary decision-history windows; action horizon one. Initialized from the pinned Qwen3-VL-4B backbone and random action head.
+
+- Training: 5000 steps, seed 42, global batch 128
+- W&B: https://wandb.ai/saberrr-zju/starvla_tasks/runs/04d16bbd
+- Formal zero-latency evaluation: pending.

zero-latency/atlantis/vla/starvla-qwenpi_v3-h1/atlantis-qwenpi_v3-hist8-zero-s42-v1/bundle.complete.json

@@ -0,0 +1,4 @@
+{
+  "checkpoint_sha256": "2f4a775690c76dc1b0e4da77a9d45f88df289e17914b895b5a65afa7fc8d214d",
+  "training_step": 5000
+}

zero-latency/atlantis/vla/starvla-qwenpi_v3-h1/atlantis-qwenpi_v3-hist8-zero-s42-v1/checkpoint.sha256

@@ -0,0 +1 @@
+2f4a775690c76dc1b0e4da77a9d45f88df289e17914b895b5a65afa7fc8d214d  checkpoints/model.pt

zero-latency/atlantis/vla/starvla-qwenpi_v3-h1/atlantis-qwenpi_v3-hist8-zero-s42-v1/checkpoints/model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f4a775690c76dc1b0e4da77a9d45f88df289e17914b895b5a65afa7fc8d214d
+size 10922725533

zero-latency/atlantis/vla/starvla-qwenpi_v3-h1/atlantis-qwenpi_v3-hist8-zero-s42-v1/collection_quality_report.json

@@ -0,0 +1,31 @@
+{
+  "action_id_counts": {
+    "0": 117589,
+    "1": 72725,
+    "2": 104979,
+    "3": 64707
+  },
+  "episodes": 200,
+  "previous_decision_admitted_fraction": 1.0,
+  "raw_frames_quantiles": [
+    7200.0,
+    7200.0,
+    7200.0,
+    7200.0,
+    7200.0
+  ],
+  "return_quantiles": [
+    36800.0,
+    36800.0,
+    37500.0,
+    42800.0,
+    42800.0
+  ],
+  "termination_reasons": {
+    "horizon": 200
+  },
+  "train_episodes": 180,
+  "train_labels": 324000,
+  "val_episodes": 20,
+  "val_labels": 36000
+}

zero-latency/atlantis/vla/starvla-qwenpi_v3-h1/atlantis-qwenpi_v3-hist8-zero-s42-v1/config.full.yaml

@@ -0,0 +1,286 @@
+framework:
+  name: QwenPI_v3
+  qwenvl:
+    base_vlm: /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-VL-4B-Instruct/snapshots/ebb281ec70b05090aa6165b016eac8ec08e71b17
+    attn_implementation: flash_attention_2
+    vl_hidden_dim: 2560
+    num_vl_layers: 36
+    enable_gradient_checkpointing: true
+  action_model:
+    action_model_type: LayerwiseFM
+    action_dim: 4
+    state_dim: 56
+    state_encoding: continuous_projector
+    task_objective: null
+    action_horizon: 1
+    repeated_diffusion_steps: 2
+    num_inference_timesteps: 4
+    add_pos_embed: true
+    max_seq_len: 1024
+    num_target_vision_tokens: 32
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    noise_s: 0.999
+    num_timestep_buckets: 1000
+    diffusion_model_cfg:
+      action_dit_hidden_dim: 1024
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      positional_embeddings: null
+      attention_head_dim: 64
+      output_dim: 1024
+      num_layers: 36
+      input_embedding_dim: 1024
+      cross_attention_dim: 1024
+      num_attention_heads: 16
+    loss_type: l1
+    future_action_window_size: 0
+    past_action_window_size: 0
+    action_env_dim: 4
+  kv_memory:
+    enabled: false
+    window: 4
+    rollout_len: 8
+    packed_train: false
+    rebased_sink: true
+datasets:
+  vla_data:
+    dataset_py: lerobot_datasets
+    include_state: true
+    data_root_dir: /home/ubuntu/codex/atari-hist8-20260924/outputs/tasks/atari_hist8_v1/data_zero/atlantis
+    data_mix: lerobot
+    eval_data_mix: lerobot__val
+    custom_mixtures_path: /home/ubuntu/codex/atari-hist8-20260924/outputs/tasks/atari_hist8_v1/data_zero/atlantis/_generated_mixtures/lerobot.json
+    action_type: discrete
+    sequential_step_sampling: false
+    eval_sequential_step_sampling: null
+    num_workers: 16
+    eval_num_workers: 16
+    prefetch_factor: 4
+    persistent_workers: true
+    pin_memory: true
+    shuffle: true
+    action_balance:
+      enabled: false
+      strategy: balanced_epoch
+      action_key: action_id
+      target_flap_fraction: 0.3
+      noop_id: 0
+      flap_id: 1
+    latency_curriculum:
+      enabled: false
+      strategy: exclusive
+      latencies: null
+      phase_steps: null
+      phase_distributions: null
+      new_latency_passes: 1.0
+      replay_passes: 0.25
+      target_total_passes: 2.0
+      final_equalization: true
+      step_budget_mode: auto
+      eval_at_phase_end: false
+      save_at_phase_end: false
+      computed_plan: null
+    per_device_batch_size: 64
+    load_all_data_for_training: true
+    num_obs_frames: 1
+    image_mode: single
+    prompt_mode: raw
+    stitch_grid:
+    - 2
+    - 2
+    obs_image_size: null
+    video_backend: torchvision_av
+dataset:
+  source_hf: latency-sensitive-bench/memory-rollouts
+  config_name: null
+  source_subdir: null
+  converted_name: atlantis_train
+  single_source_hf: latency-sensitive-bench/memory-rollouts
+  mixed_source_hf: latency-sensitive-bench/memory-rollouts
+  single_converted_name: atlantis_train
+  mixed_converted_name: atlantis_mixed_latency_train
+  single_latency_filter: null
+  mixed_latency_filter: null
+  force_download: false
+  setup_force: false
+  skip_verification: false
+  target_latency_unit: observation_steps
+  verify_rows: 200
+  max_episodes: null
+  episodes_per_latency: null
+  latency_filter: null
+  debug_subset:
+    enabled: false
+    max_episodes: 5
+    suffix: debug
+base_model:
+  repo_id: Qwen/Qwen3-VL-4B-Instruct
+initialization:
+  checkpoint_local_dir: null
+  checkpoint_hf_repo_id: null
+  checkpoint_filename: null
+trainer:
+  max_train_steps: 5000
+  num_warmup_steps: 100
+  save_interval: 500
+  eval_interval: 500
+  eval_num_batches: 50
+  per_latency_eval_num_batches: null
+  eval_action_classification: false
+  eval_action_classification_interval: null
+  cc_f1_tolerance: 1
+  learning_rate:
+    base: 3.0e-05
+    qwen_vl_interface: 3.0e-06
+    action_model: 0.0001
+  lr_scheduler_type: cosine_with_min_lr
+  scheduler_specific_kwargs:
+    min_lr_rate: 0.03333333333333333
+  freeze_modules: ''
+  freeze_vit: false
+  freeze_tied_embedding: false
+  freeze_llm_layers: []
+  loss_scale:
+    vla: 1.0
+    vlm: 0.0
+  max_grad_norm: 1.0
+  weight_decay: 0.0
+  logging_frequency: 1
+  profile_timing:
+    enabled: false
+    log_interval: 10
+  gradient_clipping: 1.0
+  gradient_accumulation_steps: 1
+  distributed_backend: deepspeed
+  is_resume: false
+  pretrained_checkpoint: null
+  resume_step: 0
+  reload_modules: null
+  optimizer:
+    name: AdamW
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 1.0e-08
+    fused: true
+  save_format: pt
+workspace_dir: WORKSPACE_DIR
+run_root_dir: /home/ubuntu/codex/atari-hist8-20260924/outputs/tasks/atari_hist8_v1/train
+seed: 42
+wandb_entity: saberrr-zju
+wandb_project: starvla_tasks
+auth:
+  env_file: null
+  hf_token_env: HF_TOKEN
+  wandb_api_key_env: WANDB_API_KEY
+paths:
+  run_root_dir: results/Checkpoints
+  dataset_local_dir: playground/Datasets/rl_games
+  dataset_cache_dir: null
+  base_model_dir: /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-VL-4B-Instruct/snapshots/ebb281ec70b05090aa6165b016eac8ec08e71b17
+  accelerate_config: starVLA/config/deepseeds/deepspeed_zero2.yaml
+launch:
+  use_accelerate: true
+  gpus: null
+  num_processes: 1
+  dry_run: false
+conda:
+  enabled: true
+  env_name: null
+rl_games:
+  model_alias: pi-0.5
+  env_eval:
+    image_size: 224
+    frameskip: 4
+    image_transform: raw_rgb
+    prompt_mode: raw
+    ghost_trail:
+      history_frames: 5
+      gamma: 1.3
+      min_alpha: 35
+      scroll_px_per_step: 4.0
+      ground_fraction: 0.22
+    seed: 42
+    fixed_episode_seeds: true
+    latency_seed_stride: 0
+    task_seed_stride: 0
+    task_description: 'Defend Atlantis by firing at descending enemies. Choose exactly
+      one action from: noop, fire, rightfire, leftfire.'
+    eval_parallel_envs: 5
+    action_chunk_execution:
+      enabled: false
+      chunk_size: null
+    atari:
+      env_id: ALE/Atlantis-v5
+      noop_max: 30
+    enabled: false
+    eval_backend: latency_bench
+    distributed_mode: none
+    vectorized:
+      enabled: false
+      batch_size: 1
+    latency:
+      prompt_map_path: null
+      mode: single
+      values:
+      - 0
+    mid_train:
+      enabled: false
+      interval_steps: 100
+      latencies:
+      - 0
+      num_episodes: 5
+      max_steps_per_episode: 3600
+    post_train:
+      enabled: false
+      latencies:
+      - 0
+      - 1
+      - 2
+      - 3
+      - 4
+      latencies_from_prompt_map: false
+      num_episodes: 5
+      max_steps_per_episode: 3600
+  task: atlantis
+  initialization_mode: scratch
+  action_carrier: native
+model: pi05
+env: atlantis
+init: scratch
+mode: single
+checkpoint:
+  load: none
+  hf_repo_id: null
+  save_best_model: false
+  save_final_model: true
+  save_pt_file: true
+  save_training_state: true
+  save_safetensors_file: false
+  local:
+    keep_last_n: 2
+  sync:
+    enabled: false
+    repo_id: null
+    keep_last_n: 0
+    sync_every_n_checkpoints: 1
+    resume_policy: local_latest
+run_id: atlantis-qwenpi_v3-hist8-zero-s42-v1
+output_dir: /home/ubuntu/codex/atari-hist8-20260924/outputs/tasks/atari_hist8_v1/train/atlantis-qwenpi_v3-hist8-zero-s42-v1
+config_yaml: /home/ubuntu/codex/atari-hist8-20260924/outputs/tasks/atari_hist8_v1/train/atlantis-qwenpi_v3-hist8-zero-s42-v1/train.yaml
+is_debug: false
+version_id: '0.21'
+wandb_name: atlantis-qwenpi_v3-hist8-zero-s42-v1
+wandb_group: decision-history-vla-h8-v2
+wandb_tags:
+- atlantis
+- qwenpi_v3
+- hist8-v2
+- history01
+- scratch
+- zero-latency
+training_latency_condition: zero-latency

zero-latency/atlantis/vla/starvla-qwenpi_v3-h1/atlantis-qwenpi_v3-hist8-zero-s42-v1/config.yaml

@@ -0,0 +1,63 @@
+framework:
+  name: QwenPI_v3
+  qwenvl:
+    base_vlm: Qwen/Qwen3-VL-4B-Instruct
+    attn_implementation: flash_attention_2
+    vl_hidden_dim: 2560
+    num_vl_layers: 36
+    enable_gradient_checkpointing: true
+  action_model:
+    action_model_type: LayerwiseFM
+    action_dim: 4
+    state_dim: 56
+    state_encoding: continuous_projector
+    task_objective: null
+    action_horizon: 1
+    repeated_diffusion_steps: 2
+    num_inference_timesteps: 4
+    add_pos_embed: true
+    max_seq_len: 1024
+    num_target_vision_tokens: 32
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    noise_s: 0.999
+    num_timestep_buckets: 1000
+    diffusion_model_cfg:
+      action_dit_hidden_dim: 1024
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      positional_embeddings: null
+      attention_head_dim: 64
+      output_dim: 1024
+      num_layers: 36
+      input_embedding_dim: 1024
+      cross_attention_dim: 1024
+      num_attention_heads: 16
+    loss_type: l1
+    future_action_window_size: 0
+    past_action_window_size: 0
+    action_env_dim: 4
+  kv_memory:
+    enabled: false
+    window: 4
+    rollout_len: 8
+    packed_train: false
+    rebased_sink: true
+version_id: '0.21'
+datasets:
+  vla_data:
+    include_state: true
+    obs_image_size: null
+    image_mode: single
+    num_obs_frames: 1
+    robot_type: rl_games_atlantis
+trainer:
+  pretrained_checkpoint: null
+backbone:
+  repo_id: Qwen/Qwen3-VL-4B-Instruct
+  revision: ebb281ec70b05090aa6165b016eac8ec08e71b17
+rl_games:
+  env_eval:
+    image_size: 224

zero-latency/atlantis/vla/starvla-qwenpi_v3-h1/atlantis-qwenpi_v3-hist8-zero-s42-v1/dataset_statistics.json

@@ -0,0 +1,400 @@
+{
+  "new_embodiment": {
+    "action": {
+      "mean": [
+        0.32630863785743713,
+        0.20220062136650085,
+        0.29162344336509705,
+        0.17986728250980377
+      ],
+      "std": [
+        0.46865805983543396,
+        0.4010893404483795,
+        0.4547502398490906,
+        0.38388657569885254
+      ],
+      "max": [
+        1.0,
+        1.0,
+        1.0,
+        1.0
+      ],
+      "min": [
+        0.0,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        0.0,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        1.0,
+        1.0,
+        1.0,
+        1.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true
+      ]
+    },
+    "state": {
+      "mean": [
+        0.995555579662323,
+        0.20145678520202637,
+        0.2900123596191406,
+        0.17871913313865662,
+        0.20145678520202637,
+        0.2900123596191406,
+        0.17871913313865662,
+        0.9961110949516296,
+        0.20156173408031464,
+        0.2902592718601227,
+        0.17883025109767914,
+        0.20156173408031464,
+        0.2902592718601227,
+        0.17883025109767914,
+        0.996666669845581,
+        0.20166049897670746,
+        0.2904568016529083,
+        0.17896604537963867,
+        0.20166049897670746,
+        0.2904568016529083,
+        0.17896604537963867,
+        0.9972222447395325,
+        0.20176543295383453,
+        0.29071295261383057,
+        0.17905555665493011,
+        0.20176543295383453,
+        0.29071295261383057,
+        0.17905555665493011,
+        0.9977777600288391,
+        0.2018703669309616,
+        0.29090431332588196,
+        0.1791728436946869,
+        0.2018703669309616,
+        0.29090431332588196,
+        0.1791728436946869,
+        0.9983333349227905,
+        0.20195679366588593,
+        0.2911142110824585,
+        0.17931482195854187,
+        0.20195679366588593,
+        0.2911142110824585,
+        0.17931482195854187,
+        0.9988889098167419,
+        0.20203395187854767,
+        0.2912592589855194,
+        0.17950616776943207,
+        0.20203395187854767,
+        0.2912592589855194,
+        0.17950616776943207,
+        0.9994444251060486,
+        0.20211727917194366,
+        0.2914629578590393,
+        0.17966975271701813,
+        0.20211727917194366,
+        0.2914629578590393,
+        0.17966975271701813
+      ],
+      "std": [
+        0.06643502414226532,
+        0.40056121349334717,
+        0.45443063974380493,
+        0.38271364569664,
+        0.40056121349334717,
+        0.45443063974380493,
+        0.38271364569664,
+        0.062142830342054256,
+        0.4006381034851074,
+        0.4545130431652069,
+        0.3828034996986389,
+        0.4006381034851074,
+        0.4545130431652069,
+        0.3828034996986389,
+        0.057567499577998636,
+        0.40071073174476624,
+        0.45459702610969543,
+        0.3829197585582733,
+        0.40071073174476624,
+        0.45459702610969543,
+        0.3829197585582733,
+        0.05258903652429528,
+        0.400748074054718,
+        0.4547581970691681,
+        0.3829844295978546,
+        0.400748074054718,
+        0.4547581970691681,
+        0.3829844295978546,
+        0.047047179192304465,
+        0.40082499384880066,
+        0.4546958804130554,
+        0.38279783725738525,
+        0.40082499384880066,
+        0.4546958804130554,
+        0.38279783725738525,
+        0.04076129198074341,
+        0.4009137749671936,
+        0.45480069518089294,
+        0.3829144835472107,
+        0.4009137749671936,
+        0.45480069518089294,
+        0.3829144835472107,
+        0.03329825401306152,
+        0.4009682536125183,
+        0.45482659339904785,
+        0.38303112983703613,
+        0.4009682536125183,
+        0.45482659339904785,
+        0.38303112983703613,
+        0.023557469248771667,
+        0.4010305106639862,
+        0.45491379499435425,
+        0.38386788964271545,
+        0.4010305106639862,
+        0.45491379499435425,
+        0.38386788964271545
+      ],
+      "max": [
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0
+      ],
+      "min": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "q01": [
+        1.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        1.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        1.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        1.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        1.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        1.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        1.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        1.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0
+      ],
+      "q99": [
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        1.0
+      ]
+    },
+    "num_transitions": 324000,
+    "num_trajectories": 180
+  }
+}

zero-latency/atlantis/vla/starvla-qwenpi_v3-h1/atlantis-qwenpi_v3-hist8-zero-s42-v1/manifest.json

@@ -0,0 +1,249 @@
+{
+  "dataset_name": "lerobot",
+  "env_name": "atlantis",
+  "episodes": 180,
+  "frames": 324000,
+  "task_prompts": [
+    "Defend Atlantis by firing at descending enemies. Choose exactly one action from: noop, fire, rightfire, leftfire."
+  ],
+  "format": "starvla_lerobot_v2_decision_v1",
+  "source": "/home/ubuntu/codex/atari-hist8-20260924/outputs/tasks/atari_hist8_v1/data_zero/atlantis/raw",
+  "robot_type": "rl_games_atlantis",
+  "action_layout": "gymnasium_discrete_v1",
+  "source_action_layout": "gymnasium_discrete_v1",
+  "action_labels": [
+    "noop",
+    "fire",
+    "rightfire",
+    "leftfire"
+  ],
+  "action_dim": 4,
+  "active_action_dim": 4,
+  "state_source": "decision_v1",
+  "state_dim": 56,
+  "state_labels": [
+    "decision_history_0",
+    "decision_history_1",
+    "decision_history_2",
+    "decision_history_3",
+    "decision_history_4",
+    "decision_history_5",
+    "decision_history_6",
+    "decision_history_7",
+    "decision_history_8",
+    "decision_history_9",
+    "decision_history_10",
+    "decision_history_11",
+    "decision_history_12",
+    "decision_history_13",
+    "decision_history_14",
+    "decision_history_15",
+    "decision_history_16",
+    "decision_history_17",
+    "decision_history_18",
+    "decision_history_19",
+    "decision_history_20",
+    "decision_history_21",
+    "decision_history_22",
+    "decision_history_23",
+    "decision_history_24",
+    "decision_history_25",
+    "decision_history_26",
+    "decision_history_27",
+    "decision_history_28",
+    "decision_history_29",
+    "decision_history_30",
+    "decision_history_31",
+    "decision_history_32",
+    "decision_history_33",
+    "decision_history_34",
+    "decision_history_35",
+    "decision_history_36",
+    "decision_history_37",
+    "decision_history_38",
+    "decision_history_39",
+    "decision_history_40",
+    "decision_history_41",
+    "decision_history_42",
+    "decision_history_43",
+    "decision_history_44",
+    "decision_history_45",
+    "decision_history_46",
+    "decision_history_47",
+    "decision_history_48",
+    "decision_history_49",
+    "decision_history_50",
+    "decision_history_51",
+    "decision_history_52",
+    "decision_history_53",
+    "decision_history_54",
+    "decision_history_55"
+  ],
+  "state_normalization": {
+    "min": [
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0,
+      0.0
+    ],
+    "max": [
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0,
+      1.0
+    ]
+  },
+  "action_history_decisions": 8,
+  "rows_unit": "decision_step",
+  "obs_fps": 15.0,
+  "obs_stride_raw_frames": 4,
+  "latency_prompt_map_path": "/home/ubuntu/codex/atari-hist8-20260924/outputs/tasks/atari_hist8_v1/data_zero/atlantis/lerobot/latency_prompt_map.json",
+  "custom_mixtures_path": "/home/ubuntu/codex/atari-hist8-20260924/outputs/tasks/atari_hist8_v1/data_zero/atlantis/_generated_mixtures/lerobot.json",
+  "integration_name": "gymnasium",
+  "gymnasium_task": {
+    "action_labels": [
+      "noop",
+      "fire",
+      "rightfire",
+      "leftfire"
+    ],
+    "action_values": [
+      0,
+      1,
+      2,
+      3
+    ],
+    "base_prompt": "Defend Atlantis by firing at descending enemies. Choose exactly one action from: noop, fire, rightfire, leftfire.",
+    "env_fps": 60,
+    "env_id": "LatencyBench/Atlantis-v0",
+    "frame_stack": 1,
+    "make_kwargs": {
+      "base_env_id": "ALE/Atlantis-v5",
+      "base_make_kwargs": {
+        "difficulty": 0,
+        "frameskip": 1,
+        "full_action_space": false,
+        "max_num_frames_per_episode": 108000,
+        "mode": 0,
+        "obs_type": "rgb",
+        "repeat_action_probability": 0.0
+      },
+      "noop_max": 30,
+      "render_mode": "rgb_array",
+      "screen_size": 84
+    },
+    "noop_action_id": 0,
+    "obs_fps": 15,
+    "registration_imports": [
+      "latency_bench.envs.gymnasium_atlantis"
+    ],
+    "task_name": "atlantis"
+  },
+  "validation_dataset_name": "lerobot__val",
+  "validation_episodes": 20,
+  "validation_frames": 36000
+}

zero-latency/atlantis/vla/starvla-qwenpi_v3-h1/atlantis-qwenpi_v3-hist8-zero-s42-v1/provenance.json

@@ -0,0 +1,33 @@
+{
+  "collection": {
+    "episodes": 200,
+    "filter_preset": "none",
+    "history": "decision_v1",
+    "history_decisions": 8,
+    "history_source": "teacher_transport",
+    "max_episode_raw_frames": 7200,
+    "rollout_overrides": {
+      "env_fps": 60,
+      "fixed_latency_ms": 0.0,
+      "latency_type": "fixed",
+      "obs_fps": 15,
+      "seed": 42
+    },
+    "student_image_frames": 1,
+    "teacher_checkpoint": "best_000026560_13598720_reward_675769.000.pth",
+    "teacher_path": "zero-latency/atlantis/small-policy/sample-factory-hist8-cpu15m-s0-v1",
+    "teacher_repo": "latency-sensitive-bench/benchmark-models",
+    "teacher_revision": "dfaeecb84cd6b5dfcc3adc0935d0ae86d0b504d2",
+    "teacher_sha256": "d41d728d8c2329d3623eb9e01bd9f3f2df4fc0d4f6f2c1f799c4badc24af78c5",
+    "train_labels": 324000,
+    "training_latency_condition": "zero"
+  },
+  "training_run_id": "atlantis-qwenpi_v3-hist8-zero-s42-v1",
+  "training_step": 5000,
+  "global_batch": 128,
+  "base_model": "Qwen/Qwen3-VL-4B-Instruct",
+  "wandb_run_id": "04d16bbd",
+  "checkpoint_sha256": "2f4a775690c76dc1b0e4da77a9d45f88df289e17914b895b5a65afa7fc8d214d",
+  "eval_protocol_version": "hist8-history01-sf-render-v4",
+  "history_encoding": "binary_0_1"
+}

zero-latency/atlantis/vla/starvla-qwenpi_v3-h1/atlantis-qwenpi_v3-hist8-zero-s42-v1/publication_status.json

@@ -0,0 +1,4 @@
+{
+  "training_status": "complete",
+  "zero_evaluation_status": "pending"
+}

zero-latency/atlantis/vla/starvla-qwenpi_v3-h1/atlantis-qwenpi_v3-hist8-zero-s42-v1/rollout_metadata.json

@@ -0,0 +1,377 @@
+{
+  "action_history_decisions": 8,
+  "action_labels": [
+    "noop",
+    "fire",
+    "rightfire",
+    "leftfire"
+  ],
+  "action_layout": "gymnasium_discrete_v1",
+  "base_prompt": "Defend Atlantis by firing at descending enemies. Choose exactly one action from: noop, fire, rightfire, leftfire.",
+  "checkpoint_config": {
+    "action_history_decisions": 8,
+    "actor_critic_share_weights": true,
+    "actor_worker_gpus": [],
+    "adam_beta1": 0.9,
+    "adam_beta2": 0.999,
+    "adam_eps": 1e-06,
+    "adaptive_stddev": false,
+    "add_latency_info": false,
+    "algo": "APPO",
+    "async_rl": true,
+    "batch_size": 1024,
+    "batched_sampling": false,
+    "benchmark": false,
+    "cli_args": {
+      "action_history_decisions": 8,
+      "adam_eps": 1e-06,
+      "adaptive_stddev": false,
+      "add_latency_info": false,
+      "algo": "APPO",
+      "async_rl": true,
+      "batch_size": 1024,
+      "device": "gpu",
+      "encoder_conv_architecture": "convnet_simple",
+      "env": "latency_gymnasium",
+      "env_fps": 60.0,
+      "episode_metrics_path": "results/checkpoints/atari_zero_hist8_cpu15m/atlantis_sf_zero_latency_hist8_cpu15m_seed0/episode_metrics.jsonl",
+      "eval_deterministic": true,
+      "eval_episodes": 5,
+      "eval_max_steps": 5000,
+      "eval_parallel_envs": 1,
+      "experiment": "atlantis_sf_zero_latency_hist8_cpu15m_seed0",
+      "experiment_summaries_interval": 1,
+      "exploration_loss": "entropy",
+      "exploration_loss_coeff": 0.003,
+      "frame_stack": 4,
+      "gae_lambda": 0.95,
+      "gamma": 0.99,
+      "gym_action_labels_json": "[\"noop\", \"fire\", \"rightfire\", \"leftfire\"]",
+      "gym_action_space_json": "{\"type\": \"discrete\", \"labels\": [\"noop\", \"fire\", \"rightfire\", \"leftfire\"], \"values\": [0, 1, 2, 3]}",
+      "gym_action_values_json": "[0, 1, 2, 3]",
+      "gym_base_prompt": "Defend Atlantis by firing at descending enemies. Choose exactly one action from: noop, fire, rightfire, leftfire.",
+      "gym_env_id": "LatencyBench/Atlantis-v0",
+      "gym_make_kwargs_json": "{\"base_env_id\": \"ALE/Atlantis-v5\", \"render_mode\": \"rgb_array\", \"screen_size\": 84, \"noop_max\": 30, \"base_make_kwargs\": {\"obs_type\": \"rgb\", \"frameskip\": 1, \"repeat_action_probability\": 0.0, \"full_action_space\": false, \"mode\": 0, \"difficulty\": 0, \"max_num_frames_per_episode\": 108000}}",
+      "gym_noop_action_id": 0,
+      "gym_noop_action_json": "\"noop\"",
+      "gym_registration_imports_json": "[\"latency_bench.envs.gymnasium_atlantis\"]",
+      "gym_task_name": "atlantis",
+      "hold_policy": "hold",
+      "keep_checkpoints": 5,
+      "latency_type": "zero",
+      "learning_rate": 0.000303,
+      "lr_schedule": "constant",
+      "max_grad_norm": 0.0,
+      "max_policy_lag": 300,
+      "mode": "train",
+      "nonlinearity": "elu",
+      "normalize_input": true,
+      "normalize_returns": true,
+      "num_batches_per_epoch": 8,
+      "num_envs_per_worker": 16,
+      "num_epochs": 2,
+      "num_policies": 1,
+      "num_workers": 4,
+      "obs_fps": 15.0,
+      "obs_scale": 1.0,
+      "ordering_policy": "issue_order_fifo",
+      "output_dir": "outputs/atari_zero_hist8_cpu15m/atlantis/train",
+      "ppo_clip_ratio": 0.1,
+      "ppo_clip_value": 0.2,
+      "recurrence": 32,
+      "restart_behavior": "resume",
+      "rollout": 64,
+      "save_every_sec": 600,
+      "seed": 0,
+      "simulator": "cpu",
+      "stats_avg": 100,
+      "train_dir": "results/checkpoints/atari_zero_hist8_cpu15m",
+      "train_for_env_steps": 15000000,
+      "use_env_info_cache": false,
+      "use_rnn": false,
+      "value_loss_coeff": 0.5,
+      "wandb_group": "atlantis",
+      "wandb_job_type": "sample_factory",
+      "wandb_project": "latency-sensitive-bench",
+      "with_vtrace": false,
+      "with_wandb": true,
+      "worker_num_splits": 1
+    },
+    "clip_reward": null,
+    "command_line": "--mode train --algo APPO --env latency_gymnasium --experiment atlantis_sf_zero_latency_hist8_cpu15m_seed0 --train_dir results/checkpoints/atari_zero_hist8_cpu15m --restart_behavior resume --device gpu --seed 0 --episode_metrics_path results/checkpoints/atari_zero_hist8_cpu15m/atlantis_sf_zero_latency_hist8_cpu15m_seed0/episode_metrics.jsonl --train_for_env_steps 15000000 --num_workers 4 --num_envs_per_worker 16 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --lr_schedule constant --nonlinearity elu --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss entropy --exploration_loss_coeff 0.003 --value_loss_coeff 0.5 --max_grad_norm 0.0 --adam_eps 1e-06 --obs_scale 1.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn False --normalize_returns True --normalize_input True --adaptive_stddev False --with_vtrace False --latency-type zero --add-latency-info False --eval-episodes 5 --eval-parallel-envs 1 --eval-max-steps 5000 --eval-deterministic True --with_wandb True --wandb_project latency-sensitive-bench --wandb_group atlantis --wandb_job_type sample_factory --encoder_conv_architecture convnet_simple --gym-task-name atlantis --gym-env-id LatencyBench/Atlantis-v0 --gym-make-kwargs-json {\"base_env_id\": \"ALE/Atlantis-v5\", \"render_mode\": \"rgb_array\", \"screen_size\": 84, \"noop_max\": 30, \"base_make_kwargs\": {\"obs_type\": \"rgb\", \"frameskip\": 1, \"repeat_action_probability\": 0.0, \"full_action_space\": false, \"mode\": 0, \"difficulty\": 0, \"max_num_frames_per_episode\": 108000}} --gym-registration-imports-json [\"latency_bench.envs.gymnasium_atlantis\"] --gym-action-space-json {\"type\": \"discrete\", \"labels\": [\"noop\", \"fire\", \"rightfire\", \"leftfire\"], \"values\": [0, 1, 2, 3]} --gym-noop-action-json \"noop\" --gym-action-labels-json [\"noop\", \"fire\", \"rightfire\", \"leftfire\"] --gym-action-values-json [0, 1, 2, 3] --gym-noop-action-id 0 --gym-base-prompt Defend Atlantis by firing at descending enemies. Choose exactly one action from: noop, fire, rightfire, leftfire. --env-fps 60 --obs-fps 15 --frame-stack 4 --action-history-decisions 8 --simulator cpu --hold-policy hold --ordering-policy issue_order_fifo --use_env_info_cache False",
+    "continuous_tanh_scale": 0.0,
+    "decoder_mlp_layers": [],
+    "decorrelate_envs_on_one_worker": true,
+    "decorrelate_experience_max_seconds": 0,
+    "default_niceness": 0,
+    "device": "gpu",
+    "encoder_conv_architecture": "convnet_simple",
+    "encoder_conv_mlp_layers": [
+      512
+    ],
+    "encoder_mlp_layers": [
+      512,
+      512
+    ],
+    "env": "latency_gymnasium",
+    "env_fps": 60,
+    "env_frameskip": 1,
+    "env_framestack": 1,
+    "env_gpu_actions": false,
+    "env_gpu_observations": true,
+    "episode_counter": false,
+    "episode_metrics_path": "results/checkpoints/atari_zero_hist8_cpu15m/atlantis_sf_zero_latency_hist8_cpu15m_seed0/episode_metrics.jsonl",
+    "eval_deterministic": true,
+    "eval_env_frameskip": 1,
+    "eval_episodes": 5,
+    "eval_latency_raw_frame_values": null,
+    "eval_max_steps": 5000,
+    "eval_parallel_envs": 1,
+    "eval_raw_reward": false,
+    "experiment": "atlantis_sf_zero_latency_hist8_cpu15m_seed0",
+    "experiment_summaries_interval": 1,
+    "exploration_loss": "entropy",
+    "exploration_loss_coeff": 0.003,
+    "export_env_raw_rgb_frames": false,
+    "fasttd3_action_chunk_horizon": 1,
+    "fasttd3_actor_action_l2": 0.0,
+    "fasttd3_compile": true,
+    "fasttd3_replay_batch_size": 32768,
+    "fasttd3_replay_capacity": 6553600,
+    "fasttd3_train_for_optimizer_steps": 10000000000,
+    "fasttd3_transitions_per_update": 64,
+    "fasttd3_v_max": 250.0,
+    "fasttd3_v_min": -250.0,
+    "fixed_latency_ms": 0.0,
+    "flush_summaries_interval": 30,
+    "force_envs_single_thread": false,
+    "frame_stack": 4,
+    "gae_lambda": 0.95,
+    "gamma": 0.99,
+    "git_hash": "unknown",
+    "git_repo_name": "not a git repository",
+    "gpu_env_batch_size": 1,
+    "gpu_env_device": "auto",
+    "gym_action_labels_json": "[\"noop\", \"fire\", \"rightfire\", \"leftfire\"]",
+    "gym_action_space_json": "{\"type\": \"discrete\", \"labels\": [\"noop\", \"fire\", \"rightfire\", \"leftfire\"], \"values\": [0, 1, 2, 3]}",
+    "gym_action_values_json": "[0, 1, 2, 3]",
+    "gym_base_prompt": "Defend Atlantis by firing at descending enemies. Choose exactly one action from: noop, fire, rightfire, leftfire.",
+    "gym_env_id": "LatencyBench/Atlantis-v0",
+    "gym_make_kwargs_json": "{\"base_env_id\": \"ALE/Atlantis-v5\", \"render_mode\": \"rgb_array\", \"screen_size\": 84, \"noop_max\": 30, \"base_make_kwargs\": {\"obs_type\": \"rgb\", \"frameskip\": 1, \"repeat_action_probability\": 0.0, \"full_action_space\": false, \"mode\": 0, \"difficulty\": 0, \"max_num_frames_per_episode\": 108000}}",
+    "gym_noop_action_id": 0,
+    "gym_noop_action_json": "\"noop\"",
+    "gym_registration_imports_json": "[\"latency_bench.envs.gymnasium_atlantis\"]",
+    "gym_state_labels_json": "",
+    "gym_task_name": "atlantis",
+    "heartbeat_interval": 20,
+    "heartbeat_reporting_interval": 180,
+    "help": false,
+    "hold_last_chunk_action": false,
+    "hold_policy": "hold",
+    "initial_model_path": null,
+    "initial_stddev": 1.0,
+    "keep_checkpoints": 5,
+    "kl_loss_coeff": 0.0,
+    "latency_profile_path": null,
+    "latency_profile_worker_slot": null,
+    "latency_seed": null,
+    "latency_type": "fixed",
+    "learning_rate": 0.000303,
+    "load_checkpoint_kind": "latest",
+    "log_to_file": true,
+    "lr_adaptive_max": 0.01,
+    "lr_adaptive_min": 1e-06,
+    "lr_schedule": "constant",
+    "lr_schedule_kl_threshold": 0.008,
+    "max_episode_steps": null,
+    "max_grad_norm": 0.0,
+    "max_latency_ms": null,
+    "max_pending_actions": null,
+    "max_policy_lag": 300,
+    "mean_latency_ms": null,
+    "min_latency_ms": null,
+    "mode": "train",
+    "nonlinearity": "elu",
+    "noop_max": null,
+    "normalize_input": true,
+    "normalize_input_keys": null,
+    "normalize_returns": true,
+    "num_batches_per_epoch": 8,
+    "num_batches_to_accumulate": 2,
+    "num_envs_per_worker": 16,
+    "num_epochs": 2,
+    "num_policies": 1,
+    "num_workers": 4,
+    "obs_fps": 15,
+    "obs_scale": 1.0,
+    "obs_subtract_mean": 0.0,
+    "optimizer": "adam",
+    "ordering_policy": "issue_order_fifo",
+    "output_dir": "outputs/atari_zero_hist8_cpu15m/atlantis/train",
+    "pbt_mix_policies_in_one_env": true,
+    "pbt_mutation_rate": 0.15,
+    "pbt_optimize_gamma": false,
+    "pbt_period_env_steps": 5000000,
+    "pbt_perturb_max": 1.5,
+    "pbt_perturb_min": 1.1,
+    "pbt_replace_fraction": 0.3,
+    "pbt_replace_reward_gap": 0.1,
+    "pbt_replace_reward_gap_absolute": 1e-06,
+    "pbt_start_mutation": 20000000,
+    "pbt_target_objective": "true_objective",
+    "pixel_format": "CHW",
+    "policy_init_gain": 1.0,
+    "policy_initialization": "orthogonal",
+    "policy_workers_per_policy": 1,
+    "ppo": null,
+    "ppo_clip_ratio": 0.1,
+    "ppo_clip_value": 0.2,
+    "recurrence": 32,
+    "restart_behavior": "resume",
+    "reward_clip": 1000.0,
+    "reward_scale": 1.0,
+    "rnn_num_layers": 1,
+    "rnn_size": 512,
+    "rnn_type": "gru",
+    "rollout": 64,
+    "save_best_after": 100000,
+    "save_best_every_sec": 5,
+    "save_best_metric": "reward",
+    "save_every_sec": 600,
+    "save_milestones_sec": -1,
+    "seed": 42,
+    "serial_mode": false,
+    "set_workers_cpu_affinity": true,
+    "shuffle_minibatches": false,
+    "simulated_worker_capacity": null,
+    "simulator": "cpu",
+    "stats_avg": 100,
+    "std_latency_ms": null,
+    "summaries_use_frameskip": true,
+    "terminal_on_life_loss": null,
+    "train_dir": "results/checkpoints/atari_zero_hist8_cpu15m",
+    "train_for_env_steps": 15000000,
+    "train_for_seconds": 10000000000,
+    "use_env_info_cache": false,
+    "use_record_episode_statistics": false,
+    "use_rnn": false,
+    "value_bootstrap": false,
+    "value_loss_coeff": 0.5,
+    "vtrace_c": 1.0,
+    "vtrace_rho": 1.0,
+    "wandb_dir": "/mnt/ssd/codex/atari-zero-hist8-20260924/wandb",
+    "wandb_group": "atlantis",
+    "wandb_job_type": "sample_factory",
+    "wandb_project": "latency-sensitive-bench",
+    "wandb_tags": null,
+    "wandb_unique_id": "atlantis_sf_zero_latency_hist8_cpu15m_seed0",
+    "wandb_user": null,
+    "with_pbt": false,
+    "with_vtrace": false,
+    "with_wandb": true,
+    "worker_num_splits": 1
+  },
+  "checkpoint_experiment_dir": "/home/ubuntu/codex/atari-hist8-20260924/outputs/tasks/atari_hist8_v1/data_zero/atlantis/teacher_runtime",
+  "checkpoint_path": "/home/ubuntu/codex/atari-hist8-20260924/outputs/tasks/atari_hist8_v1/data_zero/atlantis/teacher_runtime/best_000026560_13598720_reward_675769.000.pth",
+  "checkpoint_train_step": 26560,
+  "config_source": "/home/ubuntu/codex/atari-hist8-20260924/outputs/tasks/atari_hist8_v1/data_zero/atlantis/teacher_runtime/config.json",
+  "context_image_count": 0,
+  "context_window": 1,
+  "env_fps": 60.0,
+  "env_name": "atlantis",
+  "episode_return_field": "episode_raw_return",
+  "filter_preset": "none",
+  "filter_report": {
+    "accepted_episodes": 200,
+    "accepted_train_episodes": 180,
+    "accepted_val_episodes": 20,
+    "attempted_episodes": 200,
+    "preset": "none",
+    "rejected_episodes": 0,
+    "train_rows_after_row_filter": 324000,
+    "train_rows_before_row_filter": 324000,
+    "val_rows_after_row_filter": 36000,
+    "val_rows_before_row_filter": 36000
+  },
+  "frame_stack": 4,
+  "gymnasium_task": {
+    "action_labels": [
+      "noop",
+      "fire",
+      "rightfire",
+      "leftfire"
+    ],
+    "action_values": [
+      0,
+      1,
+      2,
+      3
+    ],
+    "base_prompt": "Defend Atlantis by firing at descending enemies. Choose exactly one action from: noop, fire, rightfire, leftfire.",
+    "env_fps": 60,
+    "env_id": "LatencyBench/Atlantis-v0",
+    "frame_stack": 4,
+    "make_kwargs": {
+      "base_env_id": "ALE/Atlantis-v5",
+      "base_make_kwargs": {
+        "difficulty": 0,
+        "frameskip": 1,
+        "full_action_space": false,
+        "max_num_frames_per_episode": 108000,
+        "mode": 0,
+        "obs_type": "rgb",
+        "repeat_action_probability": 0.0
+      },
+      "noop_max": 30,
+      "render_mode": "rgb_array",
+      "screen_size": 84
+    },
+    "noop_action_id": 0,
+    "obs_fps": 15,
+    "registration_imports": [
+      "latency_bench.envs.gymnasium_atlantis"
+    ],
+    "task_name": "atlantis"
+  },
+  "image_resolution": [
+    210,
+    160,
+    3
+  ],
+  "image_source": "env_raw_rgb",
+  "image_transform": {
+    "gamma": 1.3,
+    "ground_fraction": 0.22,
+    "history_frames": 5,
+    "image_transform": "raw_rgb",
+    "min_alpha": 35,
+    "scroll_px_per_step": 4.0
+  },
+  "image_transform_prompt_contexts": {
+    "demon_attack_ghost_trail": " You can see the recent motion of the ship, missiles, and demons in the image. Use it to make your decision.",
+    "flappy_ghost_trail": " You can see the past motion of the bird in the image. Use it to make your decision."
+  },
+  "integration_name": "gymnasium",
+  "latency_override": null,
+  "latency_source": "checkpoint",
+  "max_episodes": 200,
+  "max_steps_per_episode": 7200,
+  "max_steps_unit": "raw_frame",
+  "num_envs": 256,
+  "obs_fps": 15.0,
+  "obs_stride_raw_frames": 4,
+  "prompt_source": "sample_factory_checkpoint:gym_base_prompt",
+  "replay_num_envs": 256,
+  "reward_field": "raw_reward",
+  "reward_semantics": "raw_reward",
+  "rows_unit": "decision_step",
+  "schema_version": 2,
+  "split_seed": 0,
+  "total_episodes": 200,
+  "total_rows": 360000,
+  "train_episodes": 180,
+  "train_rows": 324000,
+  "val_episodes": 20,
+  "val_fraction": 0.1,
+  "val_rows": 36000
+}

zero-latency/atlantis/vla/starvla-qwenpi_v3-h1/atlantis-qwenpi_v3-hist8-zero-s42-v1/run_metadata.json

@@ -0,0 +1,18 @@
+{
+  "training_run_id": "atlantis-qwenpi_v3-hist8-zero-s42-v1",
+  "wandb_run_id": "04d16bbd",
+  "recipe_version": "hist8-v2",
+  "history_encoding": "binary_0_1",
+  "training_condition": "zero",
+  "initialization_seed": 42,
+  "vlm_lr": 3e-06,
+  "max_train_steps": 5000,
+  "stop_after_steps": null,
+  "num_workers_per_rank": 16,
+  "eval_num_workers_per_rank": 16,
+  "prefetch_factor": 4,
+  "base_model_revision": "ebb281ec70b05090aa6165b016eac8ec08e71b17",
+  "dataset_dir": "/home/ubuntu/codex/atari-hist8-20260924/outputs/tasks/atari_hist8_v1/data_zero/atlantis/lerobot",
+  "dataset_manifest_sha256": "212f6f231830bcc7e2730e80820a7434affb3544f4186372d61531ba629fe9a5",
+  "code_snapshot_sha256": "d6570ce4f4b12252dd870d0250c2febb4ead7950c519ad6a6ad6b70c1aceefd8"
+}