Publish latency-aware/mikasa-intercept-grab-fast/vla/starvla-qwenoft-h8/fixed2-prefix63-restart-20260925/diagnostics

latency-aware/mikasa-intercept-grab-fast/vla/starvla-qwenoft-h8/fixed2-prefix63-restart-20260925/diagnostics/controller/phase

@@ -0,0 +1 @@
+failed:1

latency-aware/mikasa-intercept-grab-fast/vla/starvla-qwenoft-h8/fixed2-prefix63-restart-20260925/diagnostics/controller/sft.log

@@ -0,0 +1,20 @@
+
+Training microbatch=64, accumulation=1, global_batch=128
+/home/ubuntu/miniconda3/envs/starvla_rl_games_openvla/bin/python -m accelerate.commands.launch --config_file /home/ubuntu/codex/mikasa-prefix63-restart-20260925/third_party/starVLA/starVLA/config/deepseeds/deepspeed_zero2.yaml --num_processes 2 --main_process_port 29704 /home/ubuntu/codex/mikasa-prefix63-restart-20260925/third_party/starVLA/starVLA/training/train_starvla.py --config_yaml /home/ubuntu/codex/mikasa-prefix63-restart-20260925/outputs/starvla/runs/mikasa_intercept_grab_fast/fixed2-prefix63-restart-20260925/sft/bootstrap.yaml
+Log: /home/ubuntu/codex/mikasa-prefix63-restart-20260925/outputs/starvla/runs/mikasa_intercept_grab_fast/fixed2-prefix63-restart-20260925/sft/bootstrap.log
+Traceback (most recent call last):
+  File "/home/ubuntu/miniconda3/envs/starvla_rl_games_openvla/lib/python3.10/runpy.py", line 196, in _run_module_as_main
+    return _run_code(code, main_globals, None,
+  File "/home/ubuntu/miniconda3/envs/starvla_rl_games_openvla/lib/python3.10/runpy.py", line 86, in _run_code
+    exec(code, run_globals)
+  File "/home/ubuntu/codex/mikasa-prefix63-restart-20260925/scripts/starvla/train_task.py", line 1143, in <module>
+    main()
+  File "/home/ubuntu/codex/mikasa-prefix63-restart-20260925/scripts/starvla/train_task.py", line 962, in main
+    trained, checkpoint = train(args, config)
+  File "/home/ubuntu/codex/mikasa-prefix63-restart-20260925/scripts/starvla/train_task.py", line 384, in train
+    run_command([
+  File "/home/ubuntu/codex/mikasa-prefix63-restart-20260925/scripts/starvla/train_task.py", line 137, in run_command
+    subprocess.run(command, cwd=cwd, env=environment() if env is None else env,
+  File "/home/ubuntu/miniconda3/envs/starvla_rl_games_openvla/lib/python3.10/subprocess.py", line 526, in run
+    raise CalledProcessError(retcode, process.args,
+subprocess.CalledProcessError: Command '['/home/ubuntu/miniconda3/envs/starvla_rl_games_openvla/bin/python', '-m', 'accelerate.commands.launch', '--config_file', '/home/ubuntu/codex/mikasa-prefix63-restart-20260925/third_party/starVLA/starVLA/config/deepseeds/deepspeed_zero2.yaml', '--num_processes', '2', '--main_process_port', '29704', '/home/ubuntu/codex/mikasa-prefix63-restart-20260925/third_party/starVLA/starVLA/training/train_starvla.py', '--config_yaml', '/home/ubuntu/codex/mikasa-prefix63-restart-20260925/outputs/starvla/runs/mikasa_intercept_grab_fast/fixed2-prefix63-restart-20260925/sft/bootstrap.yaml']' returned non-zero exit status 1.

latency-aware/mikasa-intercept-grab-fast/vla/starvla-qwenoft-h8/fixed2-prefix63-restart-20260925/diagnostics/sft/bootstrap.yaml

@@ -0,0 +1,95 @@
+run_id: bootstrap
+wandb_name: mikasa_conditioned_h8-fixed-2-qwenoft
+run_root_dir: /home/ubuntu/codex/mikasa-prefix63-restart-20260925/outputs/starvla/runs/mikasa_intercept_grab_fast/fixed2-prefix63-restart-20260925/sft/training
+seed: 0
+wandb_project: starvla_tasks
+wandb_entity: saberrr-zju
+wandb_group: vla-train
+wandb_tags:
+- mikasa_conditioned_h8
+- h8
+- formal
+- fixed-2
+training_latency_condition: fixed-2
+is_debug: false
+version_id: '0.21'
+framework:
+  name: QwenOFT
+  qwenvl:
+    base_vlm: /home/ubuntu/.cache/huggingface/hub/models--Qwen--Qwen3-VL-4B-Instruct/snapshots/ebb281ec70b05090aa6165b016eac8ec08e71b17
+    enable_gradient_checkpointing: false
+  action_model:
+    action_dim: 7
+    action_env_dim: 7
+    state_dim: 63
+    action_horizon: 8
+    future_action_window_size: 7
+    past_action_window_size: 0
+    task_objective: null
+    state_encoding: continuous_projector
+    loss_type: l1
+datasets:
+  vla_data:
+    dataset_py: latency_bench_tasks
+    lerobot_version: v3.0
+    include_state: true
+    task_contract_path: /home/ubuntu/codex/mikasa-prefix63-restart-20260925/outputs/starvla/datasets/mikasa_intercept_grab_fast/fixed2-prefix63-restart-20260925/fixed-2/lerobot/task_contract.json
+    data_root_dir: /
+    data_mix: train
+    eval_data_mix: validation
+    custom_mixtures_path: /home/ubuntu/codex/mikasa-prefix63-restart-20260925/outputs/starvla/runs/mikasa_intercept_grab_fast/fixed2-prefix63-restart-20260925/sft/bootstrap_mixture.json
+    sequential_step_sampling: true
+    per_device_batch_size: 64
+    pin_memory: true
+    persistent_workers: true
+    num_workers: 8
+    prefetch_factor: 3
+    obs_image_size:
+    - 224
+    - 224
+    video_backend: pyav
+    image_mode: single
+    stitch_grid:
+    - 1
+    - 2
+    num_obs_frames: 1
+trainer:
+  max_train_steps: 5000
+  num_warmup_steps: 100
+  save_interval: 500
+  eval_interval: 5000
+  eval_num_batches: 1
+  logging_frequency: 5
+  learning_rate:
+    base: 3.0e-05
+    action_model: 0.0001
+  lr_scheduler_type: cosine_with_min_lr
+  scheduler_specific_kwargs:
+    min_lr: 1.0e-06
+  freeze_modules: ''
+  freeze_vit: false
+  freeze_tied_embedding: false
+  freeze_llm_layers: []
+  loss_scale:
+    vla: 1.0
+  gradient_clipping: 1.0
+  gradient_accumulation_steps: 1
+  pretrained_checkpoint: null
+  is_resume: false
+  distributed_backend: deepspeed
+  optimizer:
+    name: AdamW
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 1.0e-08
+    fused: true
+  stop_after_steps: 5000
+checkpoint:
+  save_best_model: false
+  save_final_model: true
+  save_pt_file: true
+  save_training_state: true
+  local:
+    keep_last_n: 2

latency-aware/mikasa-intercept-grab-fast/vla/starvla-qwenoft-h8/fixed2-prefix63-restart-20260925/diagnostics/sft/run.json

@@ -0,0 +1,54 @@
+{
+  "task": "mikasa_conditioned_h8",
+  "model": "qwenoft",
+  "run_id": "fixed2-prefix63-restart-20260925",
+  "smoke": false,
+  "teacher": {
+    "repo_id": "latency-sensitive-bench/benchmark-models",
+    "path": "zero-latency/mikasa-intercept-grab-fast/small-policy/ppo-v1/checkpoint.pt"
+  },
+  "backbone": {
+    "repo_id": "Qwen/Qwen3-VL-4B-Instruct",
+    "allow_patterns": [
+      "*.json",
+      "*.safetensors",
+      "*.txt",
+      "*.jinja",
+      "*.model"
+    ],
+    "revision": "ebb281ec70b05090aa6165b016eac8ec08e71b17"
+  },
+  "code_revisions": {
+    "benchmark": {
+      "revision": "90b72fa0c5997f8f91c054aebaf93dfa9231794c",
+      "dirty": true
+    },
+    "starvla": {
+      "revision": "3cadac2e264b5f0013a2d10d2551abfb0a41a081",
+      "dirty": false
+    },
+    "mikasa": {
+      "revision": "017490bbb787c6efa8c025f9558e1a691965c0e0",
+      "dirty": false
+    }
+  },
+  "datasets": [
+    {
+      "repo_id": "latency-sensitive-bench/benchmark-datasets",
+      "path": "fixed-2/mikasa-intercept-grab-fast/teacher-rollouts-h8-state-v1",
+      "local_dir": "/home/ubuntu/codex/mikasa-prefix63-restart-20260925/outputs/starvla/datasets/mikasa_intercept_grab_fast/fixed2-prefix63-restart-20260925/fixed-2",
+      "revision": null,
+      "condition": "fixed-2",
+      "action_horizon": 8
+    }
+  ],
+  "seed": 0,
+  "condition": "fixed-2",
+  "no_publish": true,
+  "dagger_rounds": 10,
+  "evaluation": {
+    "seed": 4242424242,
+    "episodes": 100,
+    "every_round": false
+  }
+}