=== Avatar Training (fire-and-forget) ===
Script: train_on_hf.py
Started: 2026-03-12 23:39:04 UTC

>>> [0/4] Raising file descriptor limit...
  ulimit -n = 1048576
>>> [1/4] Setting environment...
  HF_TOKEN set: 37 chars
  WANDB_API_KEY set: 0 chars
>>> [2/4] Logging into HuggingFace...
The token has not been saved to the git credentials helper. Pass `add_to_git_credential=True` in this function directly or `--add-to-git-credential` if using via `hf`CLI if you want to set the git credential as well.
Token is valid (permission: write).
The token `beta-pc` has been saved to /root/.cache/huggingface/stored_tokens
Your token has been saved to /root/.cache/huggingface/token
Login successful.
Note: Environment variable`HF_TOKEN` is set and is the current active token independently from the token you've just configured.
⚠️  Warning: 'huggingface-cli login' is deprecated. Use 'hf auth login' instead.
>>> [3/4] Pre-downloading pretrained model...
  Downloading avatarrobotics/act-alpha-bot-right-resnet18-b12-20260312-1749...

Fetching 9 files:   0%|          | 0/9 [00:00<?, ?it/s]
Fetching 9 files:  11%|█         | 1/9 [00:00<00:01,  6.73it/s]
Fetching 9 files:  44%|████▍     | 4/9 [00:01<00:02,  2.16it/s]
Fetching 9 files: 100%|██████████| 9/9 [00:01<00:00,  5.12it/s]
  Cached at /root/.cache/huggingface/hub/models--avatarrobotics--act-alpha-bot-right-resnet18-b12-20260312-1749/snapshots/67ded6ef397ecf93b5bc9a6df4cd4666ff74c534
  GPUs: 8x NVIDIA A100-SXM4-80GB
>>> [4/4] Launching accelerate with 8 GPUs...
  Script: train_on_hf.py
  Train args: --dataset alpha_bot_right_afternoon --steps 400000 --batch-size 32 --lr 1e-4 --backbone resnet18 --preload --mixed-precision bf16 --compile --num-workers 8 --prefetch-factor 4 --save-freq 10000 --push-to-hub --pretrained-path avatarrobotics/act-alpha-bot-right-resnet18-b12-20260312-1749 

The following values were not passed to `accelerate launch` and had defaults used instead:
	`--num_machines` was set to a value of `1`
	`--mixed_precision` was set to a value of `'no'`
	`--dynamo_backend` was set to a value of `'no'`
To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`.
Model will be pushed to: avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339
Model will be pushed to: avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339
[Avatar] Mixed precision: bf16

============================================================
Training args:
  mixed_precision=bf16
  video_backend=pyav
  prefetch_factor=4
  preload=True
  compile=True
  backbone=resnet18
lerobot_train \
  --dataset.repo_id=avatarrobotics/alpha-bot-right-2026-03-11-afternoon \
  --dataset.video_backend=pyav \
  --policy.repo_id=avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339 \
  --policy.type=act \
  --policy.chunk_size=100 \
  --policy.n_action_steps=100 \
  --policy.vision_backbone=resnet18 \
  --policy.pretrained_backbone_weights=ResNet18_Weights.IMAGENET1K_V1 \
  --policy.dim_model=512 \
  --policy.n_heads=8 \
  --policy.dim_feedforward=3200 \
  --policy.n_encoder_layers=4 \
  --policy.n_decoder_layers=1 \
  --policy.use_vae=true \
  --policy.latent_dim=32 \
  --policy.n_vae_encoder_layers=4 \
  --policy.dropout=0.1 \
  --policy.kl_weight=10.0 \
  --policy.optimizer_lr=0.0001 \
  --policy.optimizer_weight_decay=1e-4 \
  --policy.optimizer_lr_backbone=0.0001 \
  --batch_size=32 \
  --steps=400000 \
  --save_freq=10000 \
  --log_freq=200 \
  --save_checkpoint=true \
  --seed=1000 \
  --num_workers=8 \
  --output_dir=outputs/train/act-vacuum-pick \
  --policy.pretrained_path=avatarrobotics/act-alpha-bot-right-resnet18-b12-20260312-1749 \
  --policy.push_to_hub=true \
  --policy.private=true
============================================================

Model will be pushed to: avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339
[Avatar] Mixed precision: bf16

============================================================
Training args:
  mixed_precision=bf16
  video_backend=pyav
  prefetch_factor=4
  preload=True
  compile=True
  backbone=resnet18
lerobot_train \
  --dataset.repo_id=avatarrobotics/alpha-bot-right-2026-03-11-afternoon \
  --dataset.video_backend=pyav \
  --policy.repo_id=avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339 \
  --policy.type=act \
  --policy.chunk_size=100 \
  --policy.n_action_steps=100 \
  --policy.vision_backbone=resnet18 \
  --policy.pretrained_backbone_weights=ResNet18_Weights.IMAGENET1K_V1 \
  --policy.dim_model=512 \
  --policy.n_heads=8 \
  --policy.dim_feedforward=3200 \
  --policy.n_encoder_layers=4 \
  --policy.n_decoder_layers=1 \
  --policy.use_vae=true \
  --policy.latent_dim=32 \
  --policy.n_vae_encoder_layers=4 \
  --policy.dropout=0.1 \
  --policy.kl_weight=10.0 \
  --policy.optimizer_lr=0.0001 \
  --policy.optimizer_weight_decay=1e-4 \
  --policy.optimizer_lr_backbone=0.0001 \
  --batch_size=32 \
  --steps=400000 \
  --save_freq=10000 \
  --log_freq=200 \
  --save_checkpoint=true \
  --seed=1000 \
  --num_workers=8 \
  --output_dir=outputs/train/act-vacuum-pick \
  --policy.pretrained_path=avatarrobotics/act-alpha-bot-right-resnet18-b12-20260312-1749 \
  --policy.push_to_hub=true \
  --policy.private=true
============================================================

Model will be pushed to: avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339
[Avatar] Mixed precision: bf16

============================================================
Training args:
  mixed_precision=bf16
  video_backend=pyav
  prefetch_factor=4
  preload=True
  compile=True
  backbone=resnet18
lerobot_train \
  --dataset.repo_id=avatarrobotics/alpha-bot-right-2026-03-11-afternoon \
  --dataset.video_backend=pyav \
  --policy.repo_id=avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339 \
  --policy.type=act \
  --policy.chunk_size=100 \
  --policy.n_action_steps=100 \
  --policy.vision_backbone=resnet18 \
  --policy.pretrained_backbone_weights=ResNet18_Weights.IMAGENET1K_V1 \
  --policy.dim_model=512 \
  --policy.n_heads=8 \
  --policy.dim_feedforward=3200 \
  --policy.n_encoder_layers=4 \
  --policy.n_decoder_layers=1 \
  --policy.use_vae=true \
  --policy.latent_dim=32 \
  --policy.n_vae_encoder_layers=4 \
  --policy.dropout=0.1 \
  --policy.kl_weight=10.0 \
  --policy.optimizer_lr=0.0001 \
  --policy.optimizer_weight_decay=1e-4 \
  --policy.optimizer_lr_backbone=0.0001 \
  --batch_size=32 \
  --steps=400000 \
  --save_freq=10000 \
  --log_freq=200 \
  --save_checkpoint=true \
  --seed=1000 \
  --num_workers=8 \
  --output_dir=outputs/train/act-vacuum-pick \
  --policy.pretrained_path=avatarrobotics/act-alpha-bot-right-resnet18-b12-20260312-1749 \
  --policy.push_to_hub=true \
  --policy.private=true
============================================================

Model will be pushed to: avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339
[Avatar] Mixed precision: bf16

============================================================
Training args:
Model will be pushed to: avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339
  mixed_precision=bf16[Avatar] Mixed precision: bf16

  video_backend=pyav
  prefetch_factor=4
  preload=True
  compile=True
  backbone=resnet18
lerobot_train \
  --dataset.repo_id=avatarrobotics/alpha-bot-right-2026-03-11-afternoon \
  --dataset.video_backend=pyav \
  --policy.repo_id=avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339 \
  --policy.type=act \
  --policy.chunk_size=100 \
  --policy.n_action_steps=100 \
  --policy.vision_backbone=resnet18 \
  --policy.pretrained_backbone_weights=ResNet18_Weights.IMAGENET1K_V1 \
  --policy.dim_model=512 \
  --policy.n_heads=8 \
  --policy.dim_feedforward=3200 \
  --policy.n_encoder_layers=4 \
  --policy.n_decoder_layers=1 \
  --policy.use_vae=true \
  --policy.latent_dim=32 \
  --policy.n_vae_encoder_layers=4 \
  --policy.dropout=0.1 \
  --policy.kl_weight=10.0 \
  --policy.optimizer_lr=0.0001 \
  --policy.optimizer_weight_decay=1e-4 \
  --policy.optimizer_lr_backbone=0.0001 \
  --batch_size=32 \
  --steps=400000 \
  --save_freq=10000 \
  --log_freq=200 \
  --save_checkpoint=true \
  --seed=1000 \
  --num_workers=8 \
  --output_dir=outputs/train/act-vacuum-pick \
  --policy.pretrained_path=avatarrobotics/act-alpha-bot-right-resnet18-b12-20260312-1749 \
  --policy.push_to_hub=true \
  --policy.private=true
============================================================

============================================================
Training args:

  mixed_precision=bf16
  video_backend=pyav
  prefetch_factor=4
  preload=True
  compile=True
  backbone=resnet18
lerobot_train \
  --dataset.repo_id=avatarrobotics/alpha-bot-right-2026-03-11-afternoon \
  --dataset.video_backend=pyav \
  --policy.repo_id=avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339 \
  --policy.type=act \
  --policy.chunk_size=100 \
  --policy.n_action_steps=100 \
  --policy.vision_backbone=resnet18 \
  --policy.pretrained_backbone_weights=ResNet18_Weights.IMAGENET1K_V1 \
  --policy.dim_model=512 \
  --policy.n_heads=8 \
  --policy.dim_feedforward=3200 \
  --policy.n_encoder_layers=4 \
  --policy.n_decoder_layers=1 \
  --policy.use_vae=true \
  --policy.latent_dim=32 \
  --policy.n_vae_encoder_layers=4 \
  --policy.dropout=0.1 \
  --policy.kl_weight=10.0 \
  --policy.optimizer_lr=0.0001 \
  --policy.optimizer_weight_decay=1e-4 \
  --policy.optimizer_lr_backbone=0.0001 \
  --batch_size=32 \
  --steps=400000 \
  --save_freq=10000 \
  --log_freq=200 \
  --save_checkpoint=true \
  --seed=1000 \
  --num_workers=8 \
  --output_dir=outputs/train/act-vacuum-pick \
  --policy.pretrained_path=avatarrobotics/act-alpha-bot-right-resnet18-b12-20260312-1749 \
  --policy.push_to_hub=true \
  --policy.private=true
============================================================

Model will be pushed to: avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339
[Avatar] Mixed precision: bf16

============================================================
Training args:
  mixed_precision=bf16
  video_backend=pyav
  prefetch_factor=4
  preload=True
  compile=True
  backbone=resnet18
lerobot_train \
  --dataset.repo_id=avatarrobotics/alpha-bot-right-2026-03-11-afternoon \
  --dataset.video_backend=pyav \
  --policy.repo_id=avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339 \
  --policy.type=act \
  --policy.chunk_size=100 \
  --policy.n_action_steps=100 \
  --policy.vision_backbone=resnet18 \
  --policy.pretrained_backbone_weights=ResNet18_Weights.IMAGENET1K_V1 \
  --policy.dim_model=512 \
  --policy.n_heads=8 \
  --policy.dim_feedforward=3200 \
  --policy.n_encoder_layers=4 \
  --policy.n_decoder_layers=1 \
  --policy.use_vae=true \
  --policy.latent_dim=32 \
  --policy.n_vae_encoder_layers=4 \
  --policy.dropout=0.1 \
  --policy.kl_weight=10.0 \
  --policy.optimizer_lr=0.0001 \
  --policy.optimizer_weight_decay=1e-4 \
  --policy.optimizer_lr_backbone=0.0001 \
  --batch_size=32 \
  --steps=400000 \
  --save_freq=10000 \
  --log_freq=200 \
  --save_checkpoint=true \
  --seed=1000 \
  --num_workers=8 \
  --output_dir=outputs/train/act-vacuum-pick \
  --policy.pretrained_path=avatarrobotics/act-alpha-bot-right-resnet18-b12-20260312-1749 \
  --policy.push_to_hub=true \
  --policy.private=true
============================================================

Model will be pushed to: avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339
[Avatar] Mixed precision: bf16

============================================================
Training args:
  mixed_precision=bf16
  video_backend=pyav
  prefetch_factor=4
  preload=True
  compile=True
  backbone=resnet18
lerobot_train \
  --dataset.repo_id=avatarrobotics/alpha-bot-right-2026-03-11-afternoon \
  --dataset.video_backend=pyav \
  --policy.repo_id=avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339 \
  --policy.type=act \
  --policy.chunk_size=100 \
  --policy.n_action_steps=100 \
  --policy.vision_backbone=resnet18 \
  --policy.pretrained_backbone_weights=ResNet18_Weights.IMAGENET1K_V1 \
  --policy.dim_model=512 \
  --policy.n_heads=8 \
  --policy.dim_feedforward=3200 \
  --policy.n_encoder_layers=4 \
  --policy.n_decoder_layers=1 \
  --policy.use_vae=true \
  --policy.latent_dim=32 \
  --policy.n_vae_encoder_layers=4 \
  --policy.dropout=0.1 \
  --policy.kl_weight=10.0 \
  --policy.optimizer_lr=0.0001 \
  --policy.optimizer_weight_decay=1e-4 \
  --policy.optimizer_lr_backbone=0.0001 \
  --batch_size=32 \
  --steps=400000 \
  --save_freq=10000 \
  --log_freq=200 \
  --save_checkpoint=true \
  --seed=1000 \
  --num_workers=8 \
  --output_dir=outputs/train/act-vacuum-pick \
  --policy.pretrained_path=avatarrobotics/act-alpha-bot-right-resnet18-b12-20260312-1749 \
  --policy.push_to_hub=true \
  --policy.private=true
============================================================

[Avatar] Mixed precision: bf16

============================================================
Training args:
  mixed_precision=bf16
  video_backend=pyav
  prefetch_factor=4
  preload=True
  compile=True
  backbone=resnet18
lerobot_train \
  --dataset.repo_id=avatarrobotics/alpha-bot-right-2026-03-11-afternoon \
  --dataset.video_backend=pyav \
  --policy.repo_id=avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339 \
  --policy.type=act \
  --policy.chunk_size=100 \
  --policy.n_action_steps=100 \
  --policy.vision_backbone=resnet18 \
  --policy.pretrained_backbone_weights=ResNet18_Weights.IMAGENET1K_V1 \
  --policy.dim_model=512 \
  --policy.n_heads=8 \
  --policy.dim_feedforward=3200 \
  --policy.n_encoder_layers=4 \
  --policy.n_decoder_layers=1 \
  --policy.use_vae=true \
  --policy.latent_dim=32 \
  --policy.n_vae_encoder_layers=4 \
  --policy.dropout=0.1 \
  --policy.kl_weight=10.0 \
  --policy.optimizer_lr=0.0001 \
  --policy.optimizer_weight_decay=1e-4 \
  --policy.optimizer_lr_backbone=0.0001 \
  --batch_size=32 \
  --steps=400000 \
  --save_freq=10000 \
  --log_freq=200 \
  --save_checkpoint=true \
  --seed=1000 \
  --num_workers=8 \
  --output_dir=outputs/train/act-vacuum-pick \
  --policy.pretrained_path=avatarrobotics/act-alpha-bot-right-resnet18-b12-20260312-1749 \
  --policy.push_to_hub=true \
  --policy.private=true
============================================================

[Avatar] cudnn.benchmark=True, tf32=True
[Avatar] cudnn.benchmark=True, tf32=True
[Avatar] cudnn.benchmark=True, tf32=True
[Avatar] cudnn.benchmark=True, tf32=True
[Avatar] cudnn.benchmark=True, tf32=True
[Avatar] cudnn.benchmark=True, tf32=True
[Avatar] cudnn.benchmark=True, tf32=True
[Avatar] cudnn.benchmark=True, tf32=True
[Avatar] Checkpoint push enabled → avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339 (with per-step tags)[Avatar] Checkpoint push enabled → avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339 (with per-step tags)

[Avatar] torch.compile enabled (will compile on first forward pass)[Avatar] torch.compile enabled (will compile on first forward pass)

[Avatar] Overriding prefetch_factor=4[Avatar] Overriding prefetch_factor=4

[Avatar] Checkpoint push enabled → avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339 (with per-step tags)
[Avatar] torch.compile enabled (will compile on first forward pass)
[Avatar] Overriding prefetch_factor=4
[Avatar] Checkpoint push enabled → avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339 (with per-step tags)
[Avatar] torch.compile enabled (will compile on first forward pass)
[Avatar] Overriding prefetch_factor=4
[Avatar] Checkpoint push enabled → avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339 (with per-step tags)
[Avatar] torch.compile enabled (will compile on first forward pass)
[Avatar] Overriding prefetch_factor=4
[Avatar] Checkpoint push enabled → avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339 (with per-step tags)
[Avatar] torch.compile enabled (will compile on first forward pass)
[Avatar] Overriding prefetch_factor=4
[Avatar] Checkpoint push enabled → avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339 (with per-step tags)
[Avatar] torch.compile enabled (will compile on first forward pass)
[Avatar] Overriding prefetch_factor=4
[Avatar] Checkpoint push enabled → avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339 (with per-step tags)
[Avatar] torch.compile enabled (will compile on first forward pass)
[Avatar] Overriding prefetch_factor=4
WARNING:lerobot.configs.policies:Device 'None' is not available. Switching to 'cuda'.
WARNING:lerobot.configs.policies:Device 'None' is not available. Switching to 'cuda'.
WARNING:lerobot.configs.policies:Device 'None' is not available. Switching to 'cuda'.
WARNING:lerobot.configs.policies:Device 'None' is not available. Switching to 'cuda'.
WARNING:lerobot.configs.policies:Device 'None' is not available. Switching to 'cuda'.
WARNING:lerobot.configs.policies:Device 'None' is not available. Switching to 'cuda'.
WARNING:lerobot.configs.policies:Device 'None' is not available. Switching to 'cuda'.
WARNING:lerobot.configs.policies:Device 'None' is not available. Switching to 'cuda'.
INFO 2026-03-12 23:39:24 ot_train.py:195 {'batch_size': 32,
 'checkpoint_path': None,
 'dataset': {'episodes': None,
             'image_transforms': {'enable': False,
                                  'max_num_transforms': 3,
                                  'random_order': False,
                                  'tfs': {'affine': {'kwargs': {'degrees': [-5.0,
                                                                            5.0],
                                                                'translate': [0.05,
                                                                              0.05]},
                                                     'type': 'RandomAffine',
                                                     'weight': 1.0},
                                          'brightness': {'kwargs': {'brightness': [0.8,
                                                                                   1.2]},
                                                         'type': 'ColorJitter',
                                                         'weight': 1.0},
                                          'contrast': {'kwargs': {'contrast': [0.8,
                                                                               1.2]},
                                                       'type': 'ColorJitter',
                                                       'weight': 1.0},
                                          'hue': {'kwargs': {'hue': [-0.05,
                                                                     0.05]},
                                                  'type': 'ColorJitter',
                                                  'weight': 1.0},
                                          'saturation': {'kwargs': {'saturation': [0.5,
                                                                                   1.5]},
                                                         'type': 'ColorJitter',
                                                         'weight': 1.0},
                                          'sharpness': {'kwargs': {'sharpness': [0.5,
                                                                                 1.5]},
                                                        'type': 'SharpnessJitter',
                                                        'weight': 1.0}}},
             'repo_id': 'avatarrobotics/alpha-bot-right-2026-03-11-afternoon',
             'revision': None,
             'root': None,
             'streaming': False,
             'use_imagenet_stats': True,
             'video_backend': 'pyav'},
 'env': None,
 'eval': {'batch_size': 50, 'n_episodes': 50, 'use_async_envs': False},
 'eval_freq': 20000,
 'job_name': 'act',
 'log_freq': 200,
 'num_workers': 8,
 'optimizer': {'betas': [0.9, 0.999],
               'eps': 1e-08,
               'grad_clip_norm': 10.0,
               'lr': 0.0001,
               'type': 'adamw',
               'weight_decay': 0.0001},
 'output_dir': 'outputs/train/act-vacuum-pick',
 'peft': None,
 'policy': {'chunk_size': 100,
            'device': 'cuda',
            'dim_feedforward': 3200,
            'dim_model': 512,
            'dropout': 0.1,
            'feedforward_activation': 'relu',
            'input_features': {},
            'kl_weight': 10.0,
            'latent_dim': 32,
            'license': None,
            'n_action_steps': 100,
            'n_decoder_layers': 1,
            'n_encoder_layers': 4,
            'n_heads': 8,
            'n_obs_steps': 1,
            'n_vae_encoder_layers': 4,
            'normalization_mapping': {'ACTION': <NormalizationMode.MEAN_STD: 'MEAN_STD'>,
                                      'STATE': <NormalizationMode.MEAN_STD: 'MEAN_STD'>,
                                      'VISUAL': <NormalizationMode.MEAN_STD: 'MEAN_STD'>},
            'optimizer_lr': 0.0001,
            'optimizer_lr_backbone': 0.0001,
            'optimizer_weight_decay': 0.0001,
            'output_features': {},
            'pre_norm': False,
            'pretrained_backbone_weights': 'ResNet18_Weights.IMAGENET1K_V1',
            'pretrained_path': 'avatarrobotics/act-alpha-bot-right-resnet18-b12-20260312-1749',
            'private': True,
            'push_to_hub': True,
            'replace_final_stride_with_dilation': False,
            'repo_id': 'avatarrobotics/act-alpha-bot-right-resnet18-b256-20260312-2339',
            'tags': None,
            'temporal_ensemble_coeff': None,
            'type': 'act',
            'use_amp': False,
            'use_peft': False,
            'use_vae': True,
            'vision_backbone': 'resnet18'},
 'rabc_epsilon': 1e-06,
 'rabc_head_mode': 'sparse',
 'rabc_kappa': 0.01,
 'rabc_progress_path': None,
 'rename_map': {},
 'resume': False,
 'save_checkpoint': True,
 'save_freq': 10000,
 'scheduler': None,
 'seed': 1000,
 'steps': 400000,
 'tolerance_s': 0.0001,
 'use_policy_training_preset': True,
 'use_rabc': False,
 'wandb': {'disable_artifact': False,
           'enable': False,
           'entity': None,
           'mode': None,
           'notes': None,
           'project': 'lerobot',
           'run_id': None}}
INFO 2026-03-12 23:39:24 ot_train.py:203 Logs will be saved locally.
INFO 2026-03-12 23:39:24 ot_train.py:215 Creating dataset

Fetching 7 files:   0%|          | 0/7 [00:00<?, ?it/s]
Fetching 7 files:  14%|█▍        | 1/7 [00:00<00:01,  3.95it/s]
Fetching 7 files:  43%|████▎     | 3/7 [00:01<00:01,  2.57it/s]
Fetching 7 files:  86%|████████▌ | 6/7 [00:04<00:00,  1.29it/s]
Fetching 7 files: 100%|██████████| 7/7 [00:04<00:00,  1.46it/s]
Fetching 7 files: 100%|██████████| 7/7 [00:04<00:00,  1.53it/s]

Fetching 947 files:   0%|          | 0/947 [00:00<?, ?it/s]
Fetching 947 files:   0%|          | 1/947 [00:00<02:49,  5.58it/s]
Fetching 947 files:   0%|          | 2/947 [00:00<06:11,  2.54it/s]
Fetching 947 files:   1%|▏         | 14/947 [00:00<00:51, 18.27it/s]
Fetching 947 files:   2%|▏         | 18/947 [00:01<01:10, 13.18it/s]
Fetching 947 files:   2%|▏         | 20/947 [00:01<01:15, 12.34it/s]
Fetching 947 files:   3%|▎         | 26/947 [00:01<00:52, 17.57it/s]
Fetching 947 files:   3%|▎         | 29/947 [00:02<01:01, 14.87it/s]
Fetching 947 files:   3%|▎         | 31/947 [00:02<01:12, 12.63it/s]
Fetching 947 files:   4%|▎         | 35/947 [00:02<01:12, 12.62it/s]
Fetching 947 files:   4%|▍         | 38/947 [00:02<01:09, 13.04it/s]
Fetching 947 files:   4%|▍         | 40/947 [00:03<01:14, 12.25it/s]
Fetching 947 files:   5%|▍         | 44/947 [00:03<01:12, 12.49it/s]
Fetching 947 files:   5%|▍         | 46/947 [00:03<01:14, 12.14it/s]
Fetching 947 files:   5%|▌         | 50/947 [00:04<01:23, 10.80it/s]
Fetching 947 files:   5%|▌         | 52/947 [00:04<01:26, 10.34it/s]
Fetching 947 files:   6%|▌         | 54/947 [00:04<01:21, 10.93it/s]
Fetching 947 files:   6%|▌         | 56/947 [00:04<01:26, 10.31it/s]
Fetching 947 files:   6%|▌         | 58/947 [00:04<01:15, 11.74it/s]
Fetching 947 files:   6%|▋         | 61/947 [00:05<01:23, 10.63it/s]
Fetching 947 files:   7%|▋         | 64/947 [00:05<01:24, 10.51it/s]
Fetching 947 files:   7%|▋         | 67/947 [00:05<01:06, 13.20it/s]
Fetching 947 files:   7%|▋         | 70/947 [00:05<00:57, 15.37it/s]
Fetching 947 files:   8%|▊         | 72/947 [00:05<01:16, 11.49it/s]
Fetching 947 files:   8%|▊         | 74/947 [00:06<01:20, 10.90it/s]
Fetching 947 files:   8%|▊         | 77/947 [00:06<01:03, 13.78it/s]
Fetching 947 files:   8%|▊         | 79/947 [00:06<01:12, 12.01it/s]
Fetching 947 files:   9%|▊         | 82/947 [00:06<01:24, 10.25it/s]
Fetching 947 files:   9%|▉         | 85/947 [00:06<01:08, 12.61it/s]
Fetching 947 files:   9%|▉         | 89/947 [00:07<01:00, 14.17it/s]
Fetching 947 files:  10%|▉         | 91/947 [00:07<01:26,  9.87it/s]
Fetching 947 files:  10%|▉         | 93/947 [00:07<01:33,  9.14it/s]
Fetching 947 files:  10%|█         | 98/947 [00:08<01:05, 12.88it/s]
Fetching 947 files:  11%|█         | 100/947 [00:08<01:10, 11.95it/s]
Fetching 947 files:  11%|█         | 102/947 [00:08<01:10, 11.92it/s]
Fetching 947 files:  11%|█         | 106/947 [00:08<01:11, 11.78it/s]
Fetching 947 files:  11%|█▏        | 108/947 [00:09<01:15, 11.16it/s]
Fetching 947 files:  12%|█▏        | 111/947 [00:09<01:07, 12.44it/s]
Fetching 947 files:  12%|█▏        | 114/947 [00:09<01:11, 11.58it/s]
Fetching 947 files:  12%|█▏        | 118/947 [00:09<01:03, 13.03it/s]
Fetching 947 files:  13%|█▎        | 120/947 [00:09<01:06, 12.51it/s]
Fetching 947 files:  13%|█▎        | 123/947 [00:10<01:10, 11.62it/s]
Fetching 947 files:  13%|█▎        | 127/947 [00:10<01:04, 12.78it/s]
Fetching 947 files:  14%|█▎        | 130/947 [00:10<01:03, 12.95it/s]
Fetching 947 files:  14%|█▍        | 132/947 [00:10<01:10, 11.61it/s]
Fetching 947 files:  14%|█▍        | 137/947 [00:11<00:49, 16.22it/s]
Fetching 947 files:  15%|█▍        | 139/947 [00:11<01:11, 11.32it/s]
Fetching 947 files:  15%|█▍        | 141/947 [00:11<01:11, 11.30it/s]
Fetching 947 files:  15%|█▌        | 145/947 [00:11<00:54, 14.63it/s]
Fetching 947 files:  16%|█▌        | 147/947 [00:12<01:06, 12.10it/s]
Fetching 947 files:  16%|█▌        | 149/947 [00:12<01:21,  9.82it/s]
Fetching 947 files:  16%|█▌        | 153/947 [00:12<00:57, 13.73it/s]
Fetching 947 files:  16%|█▋        | 155/947 [00:12<01:09, 11.43it/s]
Fetching 947 files:  17%|█▋        | 157/947 [00:13<01:24,  9.33it/s]
Fetching 947 files:  17%|█▋        | 163/947 [00:13<01:08, 11.41it/s]
Fetching 947 files:  17%|█▋        | 165/947 [00:13<01:12, 10.83it/s]
Fetching 947 files:  18%|█▊        | 169/947 [00:13<00:53, 14.58it/s]
Fetching 947 files:  18%|█▊        | 171/947 [00:14<01:18,  9.82it/s]
Fetching 947 files:  18%|█▊        | 174/947 [00:14<01:06, 11.55it/s]
Fetching 947 files:  19%|█▊        | 176/947 [00:14<01:07, 11.42it/s]
Fetching 947 files:  19%|█▉        | 178/947 [00:15<01:25,  9.04it/s]
Fetching 947 files:  19%|█▉        | 182/947 [00:15<01:04, 11.85it/s]
Fetching 947 files:  19%|█▉        | 184/947 [00:15<01:14, 10.23it/s]
Fetching 947 files:  20%|█▉        | 187/947 [00:15<01:10, 10.73it/s]
Fetching 947 files:  20%|█▉        | 189/947 [00:15<01:08, 11.12it/s]
Fetching 947 files:  20%|██        | 192/947 [00:16<00:56, 13.46it/s]
Fetching 947 files:  20%|██        | 194/947 [00:16<01:01, 12.32it/s]
Fetching 947 files:  21%|██        | 197/947 [00:16<00:58, 12.89it/s]
Fetching 947 files:  21%|██        | 199/947 [00:16<01:01, 12.11it/s]
Fetching 947 files:  21%|██        | 201/947 [00:17<01:34,  7.86it/s]
Fetching 947 files:  22%|██▏       | 208/947 [00:17<01:00, 12.31it/s]
Fetching 947 files:  22%|██▏       | 210/947 [00:17<01:05, 11.33it/s]
Fetching 947 files:  22%|██▏       | 212/947 [00:18<01:21,  9.06it/s]
Fetching 947 files:  23%|██▎       | 217/947 [00:18<00:52, 13.97it/s]
Fetching 947 files:  23%|██▎       | 220/947 [00:18<01:06, 10.99it/s]
Fetching 947 files:  24%|██▎       | 223/947 [00:18<01:01, 11.81it/s]
Fetching 947 files:  24%|██▍       | 226/947 [00:19<00:51, 13.95it/s]
Fetching 947 files:  24%|██▍       | 228/947 [00:19<00:57, 12.47it/s]
Fetching 947 files:  24%|██▍       | 230/947 [00:19<01:08, 10.52it/s]
Fetching 947 files:  24%|██▍       | 232/947 [00:19<01:06, 10.80it/s]
Fetching 947 files:  25%|██▌       | 237/947 [00:20<01:05, 10.90it/s]
Fetching 947 files:  25%|██▌       | 239/947 [00:20<01:01, 11.48it/s]
Fetching 947 files:  26%|██▌       | 243/947 [00:20<00:53, 13.24it/s]
Fetching 947 files:  26%|██▌       | 245/947 [00:20<00:53, 13.04it/s]
Fetching 947 files:  26%|██▌       | 247/947 [00:20<00:49, 14.14it/s]
Fetching 947 files:  26%|██▋       | 249/947 [00:20<00:55, 12.68it/s]
Fetching 947 files:  27%|██▋       | 253/947 [00:21<01:01, 11.30it/s]
Fetching 947 files:  27%|██▋       | 255/947 [00:21<00:56, 12.29it/s]
Fetching 947 files:  28%|██▊       | 261/947 [00:22<01:02, 10.93it/s]
Fetching 947 files:  28%|██▊       | 263/947 [00:22<01:07, 10.10it/s]
Fetching 947 files:  28%|██▊       | 268/947 [00:22<00:46, 14.53it/s]
Fetching 947 files:  29%|██▊       | 271/947 [00:22<00:54, 12.30it/s]
Fetching 947 files:  29%|██▉       | 273/947 [00:23<01:13,  9.16it/s]
Fetching 947 files:  30%|██▉       | 280/947 [00:23<00:45, 14.80it/s]
Fetching 947 files:  30%|██▉       | 283/947 [00:23<00:55, 11.87it/s]
Fetching 947 files:  30%|███       | 286/947 [00:24<01:03, 10.43it/s]
Fetching 947 files:  31%|███       | 290/947 [00:24<00:55, 11.78it/s]
Fetching 947 files:  31%|███       | 292/947 [00:24<00:56, 11.62it/s]
Fetching 947 files:  31%|███▏      | 296/947 [00:25<00:52, 12.31it/s]
Fetching 947 files:  31%|███▏      | 298/947 [00:25<01:04, 10.00it/s]
Fetching 947 files:  32%|███▏      | 302/947 [00:25<00:47, 13.55it/s]
Fetching 947 files:  32%|███▏      | 305/947 [00:25<00:49, 12.92it/s]
Fetching 947 files:  32%|███▏      | 307/947 [00:25<00:49, 13.01it/s]
Fetching 947 files:  33%|███▎      | 310/947 [00:26<00:54, 11.67it/s]
Fetching 947 files:  33%|███▎      | 313/947 [00:26<00:54, 11.64it/s]
Fetching 947 files:  33%|███▎      | 315/947 [00:26<00:52, 12.11it/s]
Fetching 947 files:  34%|███▎      | 318/947 [00:26<00:49, 12.59it/s]
Fetching 947 files:  34%|███▍      | 320/947 [00:27<00:52, 12.03it/s]
Fetching 947 files:  34%|███▍      | 322/947 [00:27<00:54, 11.39it/s]
Fetching 947 files:  34%|███▍      | 324/947 [00:27<00:51, 11.99it/s]
Fetching 947 files:  34%|███▍      | 326/947 [00:27<01:14,  8.32it/s]
Fetching 947 files:  35%|███▌      | 332/947 [00:28<00:49, 12.39it/s]
Fetching 947 files:  35%|███▌      | 334/947 [00:28<00:51, 12.00it/s]
Fetching 947 files:  36%|███▌      | 337/947 [00:28<00:49, 12.32it/s]
Fetching 947 files:  36%|███▌      | 341/947 [00:28<00:36, 16.50it/s]
Fetching 947 files:  36%|███▋      | 344/947 [00:28<00:49, 12.27it/s]
Fetching 947 files:  37%|███▋      | 346/947 [00:29<00:52, 11.36it/s]
Fetching 947 files:  37%|███▋      | 351/947 [00:29<00:42, 13.98it/s]
Fetching 947 files:  37%|███▋      | 353/947 [00:29<01:02,  9.50it/s]
Fetching 947 files:  38%|███▊      | 358/947 [00:30<00:49, 11.87it/s]
Fetching 947 files:  38%|███▊      | 360/947 [00:30<00:49, 11.76it/s]
Fetching 947 files:  38%|███▊      | 364/947 [00:30<00:47, 12.36it/s]
Fetching 947 files:  39%|███▉      | 368/947 [00:30<00:40, 14.16it/s]
Fetching 947 files:  39%|███▉      | 371/947 [00:31<00:36, 15.83it/s]
Fetching 947 files:  39%|███▉      | 373/947 [00:31<00:54, 10.63it/s]
Fetching 947 files:  40%|███▉      | 376/947 [00:31<00:58,  9.78it/s]
Fetching 947 files:  40%|████      | 380/947 [00:32<00:46, 12.14it/s]
Fetching 947 files:  41%|████      | 384/947 [00:32<00:52, 10.75it/s]
Fetching 947 files:  41%|████      | 388/947 [00:32<00:54, 10.33it/s]
Fetching 947 files:  41%|████▏     | 392/947 [00:33<00:44, 12.45it/s]
Fetching 947 files:  42%|████▏     | 394/947 [00:33<00:42, 12.87it/s]
Fetching 947 files:  42%|████▏     | 396/947 [00:33<00:58,  9.38it/s]
Fetching 947 files:  42%|████▏     | 398/947 [00:35<02:14,  4.08it/s]
Fetching 947 files:  44%|████▍     | 415/947 [00:35<00:38, 13.77it/s]
Fetching 947 files:  44%|████▍     | 419/947 [00:35<00:36, 14.36it/s]
Fetching 947 files:  45%|████▍     | 422/947 [00:35<00:39, 13.30it/s]
Fetching 947 files:  45%|████▍     | 425/947 [00:36<00:41, 12.53it/s]
Fetching 947 files:  45%|████▌     | 427/947 [00:36<00:40, 13.00it/s]
Fetching 947 files:  45%|████▌     | 430/947 [00:36<00:34, 14.86it/s]
Fetching 947 files:  46%|████▌     | 433/947 [00:36<00:45, 11.23it/s]
Fetching 947 files:  46%|████▌     | 435/947 [00:36<00:44, 11.41it/s]
Fetching 947 files:  46%|████▋     | 438/947 [00:37<00:36, 13.78it/s]
Fetching 947 files:  46%|████▋     | 440/947 [00:37<00:42, 11.95it/s]
Fetching 947 files:  47%|████▋     | 442/947 [00:37<00:47, 10.62it/s]
Fetching 947 files:  47%|████▋     | 444/947 [00:37<00:42, 11.77it/s]
Fetching 947 files:  47%|████▋     | 446/947 [00:37<00:38, 13.00it/s]
Fetching 947 files:  47%|████▋     | 449/947 [00:38<00:52,  9.40it/s]
Fetching 947 files:  48%|████▊     | 451/947 [00:38<00:47, 10.43it/s]
Fetching 947 files:  48%|████▊     | 455/947 [00:38<00:38, 12.83it/s]
Fetching 947 files:  48%|████▊     | 457/947 [00:38<00:40, 12.01it/s]
Fetching 947 files:  48%|████▊     | 459/947 [00:39<00:49,  9.84it/s]
Fetching 947 files:  49%|████▉     | 463/947 [00:39<00:39, 12.37it/s]
Fetching 947 files:  49%|████▉     | 465/947 [00:39<00:42, 11.40it/s]
Fetching 947 files:  49%|████▉     | 467/947 [00:39<00:42, 11.38it/s]
Fetching 947 files:  50%|████▉     | 470/947 [00:39<00:43, 10.87it/s]
Fetching 947 files:  50%|████▉     | 473/947 [00:40<00:41, 11.35it/s]
Fetching 947 files:  50%|█████     | 476/947 [00:40<00:34, 13.69it/s]
Fetching 947 files:  50%|█████     | 478/947 [00:40<00:45, 10.30it/s]
Fetching 947 files:  51%|█████     | 484/947 [00:40<00:33, 13.95it/s]
Fetching 947 files:  51%|█████▏    | 487/947 [00:41<00:32, 14.00it/s]
Fetching 947 files:  52%|█████▏    | 491/947 [00:41<00:32, 14.12it/s]
Fetching 947 files:  52%|█████▏    | 493/947 [00:41<00:30, 14.95it/s]
Fetching 947 files:  52%|█████▏    | 495/947 [00:41<00:31, 14.52it/s]
Fetching 947 files:  53%|█████▎    | 499/947 [00:42<00:31, 14.02it/s]
Fetching 947 files:  53%|█████▎    | 502/947 [00:42<00:27, 16.45it/s]
Fetching 947 files:  53%|█████▎    | 504/947 [00:42<00:26, 16.76it/s]
Fetching 947 files:  54%|█████▎    | 507/947 [00:42<00:32, 13.72it/s]
Fetching 947 files:  54%|█████▍    | 510/947 [00:42<00:27, 16.18it/s]
Fetching 947 files:  54%|█████▍    | 512/947 [00:42<00:26, 16.46it/s]
Fetching 947 files:  54%|█████▍    | 514/947 [00:43<00:32, 13.39it/s]
Fetching 947 files:  55%|█████▍    | 519/947 [00:43<00:44,  9.52it/s]
Fetching 947 files:  55%|█████▌    | 524/947 [00:45<01:36,  4.37it/s]
Fetching 947 files:  56%|█████▋    | 535/947 [00:46<00:48,  8.43it/s]
Fetching 947 files:  57%|█████▋    | 537/947 [00:46<01:03,  6.43it/s]
Fetching 947 files:  58%|█████▊    | 548/947 [00:50<01:34,  4.24it/s]
Fetching 947 files:  61%|██████    | 573/947 [00:57<01:36,  3.87it/s]
Fetching 947 files:  64%|██████▎   | 603/947 [00:58<00:50,  6.80it/s]
Fetching 947 files:  66%|██████▋   | 628/947 [00:59<00:32,  9.80it/s]
Fetching 947 files:  67%|██████▋   | 634/947 [01:00<00:36,  8.51it/s]
Fetching 947 files:  67%|██████▋   | 639/947 [01:03<00:51,  5.95it/s]
Fetching 947 files:  69%|██████▊   | 651/947 [01:05<00:52,  5.62it/s]
Fetching 947 files:  72%|███████▏  | 678/947 [01:09<00:42,  6.36it/s]
Fetching 947 files:  72%|███████▏  | 679/947 [01:30<03:03,  1.46it/s]
Fetching 947 files:  78%|███████▊  | 734/947 [01:40<01:13,  2.89it/s]
Fetching 947 files:  85%|████████▍ | 803/947 [02:08<00:54,  2.63it/s]
Fetching 947 files:  97%|█████████▋| 920/947 [02:30<00:07,  3.71it/s]
Fetching 947 files: 100%|██████████| 947/947 [02:30<00:00,  6.31it/s]
[Avatar] Set video tolerance_s=0.02
[Avatar] Camera transforms: arm=right, target=480x480
[Avatar] Using /dev/shm (RAM-backed, 439GB free)
[Avatar] Pre-decoding videos to JPEG: /dev/shm/avatar_frames
  Extracting 938 videos with ffmpeg...
  [100/938] 0 errors, 4s
  [200/938] 0 errors, 8s
  [300/938] 0 errors, 11s
  [400/938] 0 errors, 15s
  [500/938] 0 errors, 19s
  [600/938] 0 errors, 23s
  [700/938] 0 errors, 28s
  [800/938] 0 errors, 33s
  [900/938] 0 errors, 38s
  [938/938] 0 errors, 39s
  Done: 938/938 videos OK, 270,246 JPEGs, 18.8GB, 41s
[Avatar] Patched decode_video_frames (disk cache)
[Avatar] Preload complete (44s total)
[Avatar] Rank 0: signaled 7 other ranks
[Avatar] Installed corrupt-frame retry handler
NCCL version 2.26.2+cuda12.2
INFO 2026-03-12 23:42:44 ot_train.py:233 Creating policy
[Avatar] Set video tolerance_s=0.02
[Avatar] Set video tolerance_s=0.02
[Avatar] Camera transforms: arm=right, target=480x480
[Avatar] Rank 1: waiting for preload...
[Avatar] Rank 1: preload ready (0s)
[Avatar] Patched decode_video_frames (disk cache)
[Avatar] Installed corrupt-frame retry handler
[Avatar] Camera transforms: arm=right, target=480x480
[Avatar] Rank 5: waiting for preload...
[Avatar] Rank 5: preload ready (0s)
[Avatar] Patched decode_video_frames (disk cache)
[Avatar] Installed corrupt-frame retry handler
[Avatar] Set video tolerance_s=0.02
[Avatar] Camera transforms: arm=right, target=480x480
[Avatar] Rank 2: waiting for preload...
[Avatar] Rank 2: preload ready (0s)
[Avatar] Patched decode_video_frames (disk cache)
[Avatar] Installed corrupt-frame retry handler
[Avatar] Set video tolerance_s=0.02
[Avatar] Camera transforms: arm=right, target=480x480
[Avatar] Rank 7: waiting for preload...
[Avatar] Rank 7: preload ready (0s)
[Avatar] Patched decode_video_frames (disk cache)
[Avatar] Installed corrupt-frame retry handler
[Avatar] Set video tolerance_s=0.02
[Avatar] Camera transforms: arm=right, target=480x480
[Avatar] Rank 3: waiting for preload...
[Avatar] Rank 3: preload ready (0s)
[Avatar] Patched decode_video_frames (disk cache)
[Avatar] Installed corrupt-frame retry handler
[Avatar] Set video tolerance_s=0.02
[Avatar] Camera transforms: arm=right, target=480x480
[Avatar] Rank 4: waiting for preload...
[Avatar] Rank 4: preload ready (0s)
[Avatar] Patched decode_video_frames (disk cache)
[Avatar] Installed corrupt-frame retry handler
[Avatar] Set video tolerance_s=0.02
[Avatar] Camera transforms: arm=right, target=480x480
[Avatar] Rank 6: waiting for preload...
[Avatar] Rank 6: preload ready (0s)
[Avatar] Patched decode_video_frames (disk cache)
[Avatar] Installed corrupt-frame retry handler
Downloading: "https://download.pytorch.org/models/resnet18-f37072fd.pth" to /root/.cache/torch/hub/checkpoints/resnet18-f37072fd.pth
Downloading: "https://download.pytorch.org/models/resnet18-f37072fd.pth" to /root/.cache/torch/hub/checkpoints/resnet18-f37072fd.pth
Downloading: "https://download.pytorch.org/models/resnet18-f37072fd.pth" to /root/.cache/torch/hub/checkpoints/resnet18-f37072fd.pth
Downloading: "https://download.pytorch.org/models/resnet18-f37072fd.pth" to /root/.cache/torch/hub/checkpoints/resnet18-f37072fd.pth
Downloading: "https://download.pytorch.org/models/resnet18-f37072fd.pth" to /root/.cache/torch/hub/checkpoints/resnet18-f37072fd.pth
Downloading: "https://download.pytorch.org/models/resnet18-f37072fd.pth" to /root/.cache/torch/hub/checkpoints/resnet18-f37072fd.pth
Downloading: "https://download.pytorch.org/models/resnet18-f37072fd.pth" to /root/.cache/torch/hub/checkpoints/resnet18-f37072fd.pth
Downloading: "https://download.pytorch.org/models/resnet18-f37072fd.pth" to /root/.cache/torch/hub/checkpoints/resnet18-f37072fd.pth

  0%|          | 0.00/44.7M [00:00<?, ?B/s]
  0%|          | 0.00/44.7M [00:00<?, ?B/s]
  0%|          | 0.00/44.7M [00:00<?, ?B/s]
 15%|█▌        | 6.75M/44.7M [00:00<00:00, 46.8MB/s]
 64%|██████▍   | 28.8M/44.7M [00:00<00:00, 101MB/s] 
 19%|█▉        | 8.50M/44.7M [00:00<00:01, 28.8MB/s]
100%|██████████| 44.7M/44.7M [00:00<00:00, 126MB/s]

 39%|███▉      | 17.5M/44.7M [00:00<00:00, 48.0MB/s]
 58%|█████▊    | 26.0M/44.7M [00:00<00:00, 53.0MB/s]
 76%|███████▌  | 33.9M/44.7M [00:00<00:00, 58.3MB/s]
 15%|█▌        | 6.88M/44.7M [00:00<00:04, 9.60MB/s]
100%|██████████| 44.7M/44.7M [00:00<00:00, 62.8MB/s]

 24%|██▎       | 10.5M/44.7M [00:00<00:02, 13.1MB/s]
  0%|          | 0.00/44.7M [00:00<?, ?B/s]
  5%|▌         | 2.25M/44.7M [00:00<00:01, 22.9MB/s]
 39%|███▉      | 17.6M/44.7M [00:01<00:01, 23.3MB/s][Avatar] Compiling policy with torch.compile (mode=reduce-overhead)...

 11%|█         | 4.75M/44.7M [00:00<00:01, 24.7MB/s]
 48%|████▊     | 21.5M/44.7M [00:01<00:01, 18.9MB/s]
 64%|██████▎   | 28.4M/44.7M [00:01<00:00, 27.0MB/s][Avatar] Compiling policy with torch.compile (mode=reduce-overhead)...

 72%|███████▏  | 32.4M/44.7M [00:01<00:00, 23.1MB/s]
 80%|███████▉  | 35.6M/44.7M [00:01<00:00, 21.9MB/s][Avatar] Policy compiled (first step will be slow)

 16%|█▌        | 7.12M/44.7M [00:01<00:07, 5.59MB/s]
 22%|██▏       | 9.62M/44.7M [00:01<00:04, 8.20MB/s]
 27%|██▋       | 11.9M/44.7M [00:01<00:04, 7.63MB/s][Avatar] Policy compiled (first step will be slow)

 33%|███▎      | 14.9M/44.7M [00:01<00:02, 10.7MB/s]
 86%|████████▌ | 38.4M/44.7M [00:02<00:00, 10.8MB/s]
 39%|███▉      | 17.6M/44.7M [00:01<00:02, 13.4MB/s]
 45%|████▌     | 20.1M/44.7M [00:01<00:01, 13.8MB/s]
 53%|█████▎    | 23.5M/44.7M [00:02<00:01, 17.8MB/s]
 59%|█████▉    | 26.4M/44.7M [00:02<00:01, 18.6MB/s]
 67%|██████▋   | 30.1M/44.7M [00:02<00:00, 23.1MB/s]
 82%|████████▏ | 36.8M/44.7M [00:02<00:00, 33.0MB/s]
 90%|█████████ | 40.4M/44.7M [00:03<00:00, 6.67MB/s]
100%|██████████| 44.7M/44.7M [00:03<00:00, 13.5MB/s]

 91%|█████████ | 40.5M/44.7M [00:02<00:00, 21.0MB/s]
  0%|          | 0.00/44.7M [00:00<?, ?B/s]
  0%|          | 0.00/44.7M [00:00<?, ?B/s]
 97%|█████████▋| 43.4M/44.7M [00:03<00:00, 16.2MB/s]
  1%|          | 512k/44.7M [00:00<00:09, 4.68MB/s]
100%|██████████| 44.7M/44.7M [00:03<00:00, 14.3MB/s]

  2%|▏         | 896k/44.7M [00:00<00:16, 2.83MB/s]
  2%|▏         | 1.00M/44.7M [00:00<00:22, 2.02MB/s]
  3%|▎         | 1.38M/44.7M [00:00<00:27, 1.67MB/s]
  6%|▋         | 2.88M/44.7M [00:00<00:10, 4.33MB/s][Avatar] Compiling policy with torch.compile (mode=reduce-overhead)...

 12%|█▏        | 5.25M/44.7M [00:00<00:04, 8.67MB/s]
 17%|█▋        | 7.75M/44.7M [00:01<00:03, 12.7MB/s]
  0%|          | 0.00/44.7M [00:00<?, ?B/s]
  0%|          | 128k/44.7M [00:00<00:41, 1.11MB/s]
 21%|██▏       | 9.50M/44.7M [00:01<00:02, 12.4MB/s]
 11%|█         | 5.00M/44.7M [00:00<00:01, 27.5MB/s]
  3%|▎         | 1.25M/44.7M [00:01<00:56, 807kB/s]
  4%|▍         | 1.75M/44.7M [00:01<00:36, 1.24MB/s]
 19%|█▉        | 8.38M/44.7M [00:00<00:02, 14.8MB/s][Avatar] Policy compiled (first step will be slow)

 25%|██▍       | 11.0M/44.7M [00:01<00:05, 6.67MB/s]
 35%|███▍      | 15.5M/44.7M [00:00<00:01, 25.7MB/s]
 27%|██▋       | 12.1M/44.7M [00:01<00:04, 7.34MB/s]
 30%|███       | 13.5M/44.7M [00:01<00:03, 8.40MB/s]
 33%|███▎      | 14.8M/44.7M [00:02<00:03, 8.35MB/s]
 36%|███▌      | 16.0M/44.7M [00:02<00:03, 8.51MB/s]
 40%|████      | 17.9M/44.7M [00:02<00:02, 9.68MB/s]
 43%|████▎     | 19.2M/44.7M [00:02<00:02, 10.6MB/s]
 52%|█████▏    | 23.4M/44.7M [00:02<00:01, 13.3MB/s]
 58%|█████▊    | 26.1M/44.7M [00:02<00:01, 16.3MB/s]
 62%|██████▏   | 27.9M/44.7M [00:03<00:01, 15.9MB/s]
 67%|██████▋   | 30.0M/44.7M [00:03<00:00, 17.3MB/s]
 71%|███████▏  | 31.9M/44.7M [00:03<00:00, 16.0MB/s]
 75%|███████▌  | 33.5M/44.7M [00:03<00:00, 15.0MB/s]
 79%|███████▉  | 35.2M/44.7M [00:03<00:00, 15.7MB/s]
 83%|████████▎ | 37.1M/44.7M [00:03<00:00, 16.7MB/s]
 87%|████████▋ | 38.9M/44.7M [00:03<00:00, 15.8MB/s]
 91%|█████████ | 40.5M/44.7M [00:03<00:00, 16.0MB/s]
 95%|█████████▍| 42.2M/44.7M [00:03<00:00, 16.6MB/s]
 98%|█████████▊| 43.9M/44.7M [00:04<00:00, 15.2MB/s]
100%|██████████| 44.7M/44.7M [00:04<00:00, 11.4MB/s]
[Avatar] Compiling policy with torch.compile (mode=reduce-overhead)...
[Avatar] Compiling policy with torch.compile (mode=reduce-overhead)...
[Avatar] Policy compiled (first step will be slow)

  5%|▍         | 2.12M/44.7M [00:06<03:26, 216kB/s] 
  6%|▌         | 2.62M/44.7M [00:06<02:11, 334kB/s][Avatar] Policy compiled (first step will be slow)

  8%|▊         | 3.38M/44.7M [00:06<01:14, 578kB/s]
 10%|▉         | 4.25M/44.7M [00:06<00:44, 960kB/s]
 13%|█▎        | 5.88M/44.7M [00:07<00:21, 1.92MB/s]
 17%|█▋        | 7.62M/44.7M [00:07<00:12, 3.20MB/s]
 20%|█▉        | 8.88M/44.7M [00:07<00:11, 3.21MB/s]
 22%|██▏       | 9.88M/44.7M [00:07<00:11, 3.27MB/s]
 27%|██▋       | 12.0M/44.7M [00:08<00:06, 5.26MB/s]
 42%|████▏     | 18.6M/44.7M [00:07<00:15, 1.80MB/s]
 30%|██▉       | 13.2M/44.7M [00:08<00:05, 5.57MB/s]
 49%|████▉     | 21.9M/44.7M [00:07<00:09, 2.48MB/s]
 55%|█████▍    | 24.4M/44.7M [00:07<00:06, 3.15MB/s]
 60%|█████▉    | 26.8M/44.7M [00:07<00:04, 3.96MB/s]
 32%|███▏      | 14.2M/44.7M [00:08<00:06, 4.67MB/s]
 35%|███▍      | 15.6M/44.7M [00:08<00:05, 5.54MB/s]
 37%|███▋      | 16.5M/44.7M [00:08<00:06, 4.72MB/s]
 39%|███▊      | 17.2M/44.7M [00:09<00:07, 4.08MB/s]
 65%|██████▍   | 29.0M/44.7M [00:08<00:04, 3.54MB/s]
 46%|████▌     | 20.6M/44.7M [00:09<00:03, 8.01MB/s]
 50%|█████     | 22.4M/44.7M [00:09<00:02, 9.56MB/s]
 69%|██████▉   | 30.9M/44.7M [00:08<00:03, 4.07MB/s]
 80%|████████  | 35.8M/44.7M [00:08<00:01, 7.02MB/s]
 88%|████████▊ | 39.1M/44.7M [00:08<00:00, 9.13MB/s]
  0%|          | 0.00/44.7M [00:00<?, ?B/s]
  3%|▎         | 1.50M/44.7M [00:00<00:02, 15.7MB/s]
 53%|█████▎    | 23.8M/44.7M [00:10<00:05, 4.09MB/s]
 58%|█████▊    | 25.8M/44.7M [00:10<00:03, 5.64MB/s]
 60%|██████    | 27.0M/44.7M [00:10<00:02, 6.50MB/s]
 93%|█████████▎| 41.5M/44.7M [00:09<00:00, 5.74MB/s]
100%|██████████| 44.7M/44.7M [00:09<00:00, 4.77MB/s]

 66%|██████▋   | 29.6M/44.7M [00:11<00:02, 6.81MB/s]
 12%|█▏        | 5.38M/44.7M [00:01<00:13, 2.98MB/s]
 69%|██████▉   | 30.8M/44.7M [00:14<00:09, 1.53MB/s]
 71%|███████   | 31.6M/44.7M [00:14<00:08, 1.68MB/s]
 72%|███████▏  | 32.2M/44.7M [00:15<00:08, 1.58MB/s]
 73%|███████▎  | 32.8M/44.7M [00:15<00:07, 1.73MB/s][Avatar] Compiling policy with torch.compile (mode=reduce-overhead)...

 74%|███████▍  | 33.2M/44.7M [00:16<00:10, 1.13MB/s]
 75%|███████▌  | 33.6M/44.7M [00:16<00:10, 1.13MB/s][Avatar] Policy compiled (first step will be slow)

 76%|███████▋  | 34.1M/44.7M [00:17<00:09, 1.17MB/s]
 79%|███████▉  | 35.2M/44.7M [00:17<00:05, 1.90MB/s]
 83%|████████▎ | 37.0M/44.7M [00:17<00:02, 3.25MB/s]
 87%|████████▋ | 39.0M/44.7M [00:17<00:01, 5.00MB/s]
 92%|█████████▏| 40.9M/44.7M [00:17<00:00, 6.89MB/s]
 98%|█████████▊| 43.6M/44.7M [00:17<00:00, 10.2MB/s]
100%|██████████| 44.7M/44.7M [00:17<00:00, 2.64MB/s]
[Avatar] Compiling policy with torch.compile (mode=reduce-overhead)...
[Avatar] Policy compiled (first step will be slow)

 15%|█▍        | 6.62M/44.7M [00:15<00:13, 2.98MB/s]
 15%|█▌        | 6.75M/44.7M [03:32<27:44, 23.9kB/s]
 17%|█▋        | 7.38M/44.7M [03:32<23:02, 28.3kB/s]
 18%|█▊        | 7.88M/44.7M [03:45<22:43, 28.3kB/s]
 18%|█▊        | 8.00M/44.7M [04:14<26:02, 24.6kB/s]
 18%|█▊        | 8.25M/44.7M [04:14<23:26, 27.1kB/s]
 20%|█▉        | 8.75M/44.7M [04:14<18:16, 34.3kB/s]
 20%|██        | 9.12M/44.7M [04:15<14:47, 42.0kB/s]
 21%|██▏       | 9.50M/44.7M [04:15<11:38, 52.8kB/s]
 21%|██▏       | 9.50M/44.7M [04:30<11:38, 52.8kB/s]
 22%|██▏       | 9.62M/44.7M [04:36<19:29, 31.4kB/s]
 22%|██▏       | 9.62M/44.7M [04:50<19:29, 31.4kB/s][rank6]:[E312 23:52:47.560128968 ProcessGroupNCCL.cpp:632] [Rank 6] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=2, OpType=ALLREDUCE, NumelIn=1, NumelOut=1, Timeout(ms)=600000) ran for 600082 milliseconds before timing out.
[rank6]:[E312 23:52:47.560612504 ProcessGroupNCCL.cpp:2271] [PG ID 0 PG GUID 0(default_pg) Rank 6]  failure detected by watchdog at work sequence id: 2 PG status: last enqueued work: 2, last completed work: 1
[rank6]:[E312 23:52:47.560626339 ProcessGroupNCCL.cpp:670] Stack trace of the failed collective not found, potentially because FlightRecorder is disabled. You can enable it by setting TORCH_NCCL_TRACE_BUFFER_SIZE to a non-zero value.
[rank6]:[E312 23:52:47.560674566 ProcessGroupNCCL.cpp:2106] [PG ID 0 PG GUID 0(default_pg) Rank 6] First PG on this rank to signal dumping.
[rank2]:[E312 23:52:47.027303888 ProcessGroupNCCL.cpp:632] [Rank 2] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=2, OpType=ALLREDUCE, NumelIn=1, NumelOut=1, Timeout(ms)=600000) ran for 600048 milliseconds before timing out.
[rank2]:[E312 23:52:47.027753041 ProcessGroupNCCL.cpp:2271] [PG ID 0 PG GUID 0(default_pg) Rank 2]  failure detected by watchdog at work sequence id: 2 PG status: last enqueued work: 2, last completed work: 1
[rank2]:[E312 23:52:47.027765739 ProcessGroupNCCL.cpp:670] Stack trace of the failed collective not found, potentially because FlightRecorder is disabled. You can enable it by setting TORCH_NCCL_TRACE_BUFFER_SIZE to a non-zero value.
[rank2]:[E312 23:52:47.027820756 ProcessGroupNCCL.cpp:2106] [PG ID 0 PG GUID 0(default_pg) Rank 2] First PG on this rank to signal dumping.
[rank0]:[E312 23:52:47.125792343 ProcessGroupNCCL.cpp:1685] [PG ID 0 PG GUID 0(default_pg) Rank 0] Observed flight recorder dump signal from another rank via TCPStore.
[rank0]:[E312 23:52:47.125988125 ProcessGroupNCCL.cpp:1746] [PG ID 0 PG GUID 0(default_pg) Rank 0] Received a dump signal due to a collective timeout from  rank 2 and we will try our best to dump the debug info. Last enqueued NCCL work: 2, last completed NCCL work: 1.This is most likely caused by incorrect usages of collectives, e.g., wrong sizes used across ranks, the order of collectives is not same for all ranks or the scheduled collective, for some reason, didn't run. Additionally, this can be caused by GIL deadlock or other reasons such as network errors or bugs in the communications library (e.g. NCCL), etc. 
[rank0]:[E312 23:52:47.126227440 ProcessGroupNCCL.cpp:1536] [PG ID 0 PG GUID 0(default_pg) Rank 0] ProcessGroupNCCL preparing to dump debug info. Include stack trace: 1
[rank2]:[E312 23:52:48.331824803 ProcessGroupNCCL.cpp:1746] [PG ID 0 PG GUID 0(default_pg) Rank 2] Received a dump signal due to a collective timeout from this local rank and we will try our best to dump the debug info. Last enqueued NCCL work: 2, last completed NCCL work: 1.This is most likely caused by incorrect usages of collectives, e.g., wrong sizes used across ranks, the order of collectives is not same for all ranks or the scheduled collective, for some reason, didn't run. Additionally, this can be caused by GIL deadlock or other reasons such as network errors or bugs in the communications library (e.g. NCCL), etc. 
[rank5]:[E312 23:52:48.331852659 ProcessGroupNCCL.cpp:1685] [PG ID 0 PG GUID 0(default_pg) Rank 5] Observed flight recorder dump signal from another rank via TCPStore.
[rank7]:[E312 23:52:48.331951518 ProcessGroupNCCL.cpp:1685] [PG ID 0 PG GUID 0(default_pg) Rank 7] Observed flight recorder dump signal from another rank via TCPStore.
[rank3]:[E312 23:52:48.331961172 ProcessGroupNCCL.cpp:1685] [PG ID 0 PG GUID 0(default_pg) Rank 3] Observed flight recorder dump signal from another rank via TCPStore.
[rank2]:[E312 23:52:48.332188719 ProcessGroupNCCL.cpp:1536] [PG ID 0 PG GUID 0(default_pg) Rank 2] ProcessGroupNCCL preparing to dump debug info. Include stack trace: 1
[rank7]:[E312 23:52:48.332206972 ProcessGroupNCCL.cpp:1746] [PG ID 0 PG GUID 0(default_pg) Rank 7] Received a dump signal due to a collective timeout from  rank 2 and we will try our best to dump the debug info. Last enqueued NCCL work: 2, last completed NCCL work: 1.This is most likely caused by incorrect usages of collectives, e.g., wrong sizes used across ranks, the order of collectives is not same for all ranks or the scheduled collective, for some reason, didn't run. Additionally, this can be caused by GIL deadlock or other reasons such as network errors or bugs in the communications library (e.g. NCCL), etc. 
[rank5]:[E312 23:52:48.332267062 ProcessGroupNCCL.cpp:1746] [PG ID 0 PG GUID 0(default_pg) Rank 5] Received a dump signal due to a collective timeout from  rank 2 and we will try our best to dump the debug info. Last enqueued NCCL work: 1, last completed NCCL work: 1.This is most likely caused by incorrect usages of collectives, e.g., wrong sizes used across ranks, the order of collectives is not same for all ranks or the scheduled collective, for some reason, didn't run. Additionally, this can be caused by GIL deadlock or other reasons such as network errors or bugs in the communications library (e.g. NCCL), etc. 
[rank3]:[E312 23:52:48.332276814 ProcessGroupNCCL.cpp:1746] [PG ID 0 PG GUID 0(default_pg) Rank 3] Received a dump signal due to a collective timeout from  rank 2 and we will try our best to dump the debug info. Last enqueued NCCL work: 2, last completed NCCL work: 1.This is most likely caused by incorrect usages of collectives, e.g., wrong sizes used across ranks, the order of collectives is not same for all ranks or the scheduled collective, for some reason, didn't run. Additionally, this can be caused by GIL deadlock or other reasons such as network errors or bugs in the communications library (e.g. NCCL), etc. 
[rank7]:[E312 23:52:48.332386105 ProcessGroupNCCL.cpp:1536] [PG ID 0 PG GUID 0(default_pg) Rank 7] ProcessGroupNCCL preparing to dump debug info. Include stack trace: 1
[rank3]:[E312 23:52:48.332513600 ProcessGroupNCCL.cpp:1536] [PG ID 0 PG GUID 0(default_pg) Rank 3] ProcessGroupNCCL preparing to dump debug info. Include stack trace: 1
[rank5]:[E312 23:52:48.332525100 ProcessGroupNCCL.cpp:1536] [PG ID 0 PG GUID 0(default_pg) Rank 5] ProcessGroupNCCL preparing to dump debug info. Include stack trace: 1
[rank6]:[E312 23:52:48.334694742 ProcessGroupNCCL.cpp:1746] [PG ID 0 PG GUID 0(default_pg) Rank 6] Received a dump signal due to a collective timeout from this local rank and we will try our best to dump the debug info. Last enqueued NCCL work: 2, last completed NCCL work: 1.This is most likely caused by incorrect usages of collectives, e.g., wrong sizes used across ranks, the order of collectives is not same for all ranks or the scheduled collective, for some reason, didn't run. Additionally, this can be caused by GIL deadlock or other reasons such as network errors or bugs in the communications library (e.g. NCCL), etc. 
[rank1]:[E312 23:52:48.334735312 ProcessGroupNCCL.cpp:1685] [PG ID 0 PG GUID 0(default_pg) Rank 1] Observed flight recorder dump signal from another rank via TCPStore.
[rank4]:[E312 23:52:48.334809414 ProcessGroupNCCL.cpp:1685] [PG ID 0 PG GUID 0(default_pg) Rank 4] Observed flight recorder dump signal from another rank via TCPStore.
[rank1]:[E312 23:52:48.334879435 ProcessGroupNCCL.cpp:1746] [PG ID 0 PG GUID 0(default_pg) Rank 1] Received a dump signal due to a collective timeout from  rank 2 and we will try our best to dump the debug info. Last enqueued NCCL work: 2, last completed NCCL work: 1.This is most likely caused by incorrect usages of collectives, e.g., wrong sizes used across ranks, the order of collectives is not same for all ranks or the scheduled collective, for some reason, didn't run. Additionally, this can be caused by GIL deadlock or other reasons such as network errors or bugs in the communications library (e.g. NCCL), etc. 
[rank6]:[E312 23:52:48.334872716 ProcessGroupNCCL.cpp:1536] [PG ID 0 PG GUID 0(default_pg) Rank 6] ProcessGroupNCCL preparing to dump debug info. Include stack trace: 1
[rank4]:[E312 23:52:48.335026171 ProcessGroupNCCL.cpp:1746] [PG ID 0 PG GUID 0(default_pg) Rank 4] Received a dump signal due to a collective timeout from  rank 2 and we will try our best to dump the debug info. Last enqueued NCCL work: 2, last completed NCCL work: 1.This is most likely caused by incorrect usages of collectives, e.g., wrong sizes used across ranks, the order of collectives is not same for all ranks or the scheduled collective, for some reason, didn't run. Additionally, this can be caused by GIL deadlock or other reasons such as network errors or bugs in the communications library (e.g. NCCL), etc. 
[rank1]:[E312 23:52:48.335115822 ProcessGroupNCCL.cpp:1536] [PG ID 0 PG GUID 0(default_pg) Rank 1] ProcessGroupNCCL preparing to dump debug info. Include stack trace: 1
[rank4]:[E312 23:52:48.335325252 ProcessGroupNCCL.cpp:1536] [PG ID 0 PG GUID 0(default_pg) Rank 4] ProcessGroupNCCL preparing to dump debug info. Include stack trace: 1
INFO 2026-03-12 23:52:48 ot_train.py:288 Creating optimizer and scheduler
INFO 2026-03-12 23:52:48 ot_train.py:323 Output dir: outputs/train/act-vacuum-pick
INFO 2026-03-12 23:52:48 ot_train.py:330 cfg.steps=400000 (400K)
INFO 2026-03-12 23:52:48 ot_train.py:331 dataset.num_frames=134890 (135K)
INFO 2026-03-12 23:52:48 ot_train.py:332 dataset.num_episodes=469
INFO 2026-03-12 23:52:48 ot_train.py:335 Effective batch size: 32 x 8 = 256
INFO 2026-03-12 23:52:48 ot_train.py:336 num_learnable_params=51601288 (52M)
INFO 2026-03-12 23:52:48 ot_train.py:337 num_total_params=51601288 (52M)
[rank2]:[E312 23:52:48.861853434 ProcessGroupNCCL.cpp:684] [Rank 2] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data.
[rank2]:[E312 23:52:48.861876278 ProcessGroupNCCL.cpp:698] [Rank 2] To avoid data inconsistency, we are taking the entire process down.
[rank2]:[E312 23:52:48.863210237 ProcessGroupNCCL.cpp:1899] [PG ID 0 PG GUID 0(default_pg) Rank 2] Process group watchdog thread terminated with exception: [Rank 2] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=2, OpType=ALLREDUCE, NumelIn=1, NumelOut=1, Timeout(ms)=600000) ran for 600048 milliseconds before timing out.
Exception raised from checkTimeout at /pytorch/torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:635 (most recent call first):
frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x790fc69785e8 in /usr/local/lib/python3.11/dist-packages/torch/lib/libc10.so)
frame #1: c10d::ProcessGroupNCCL::WorkNCCL::checkTimeout(std::optional<std::chrono::duration<long, std::ratio<1l, 1000l> > >) + 0x23d (0x790f713cba6d in /usr/local/lib/python3.11/dist-packages/torch/lib/libtorch_cuda.so)
frame #2: c10d::ProcessGroupNCCL::watchdogHandler() + 0xc80 (0x790f713cd7f0 in /usr/local/lib/python3.11/dist-packages/torch/lib/libtorch_cuda.so)
frame #3: c10d::ProcessGroupNCCL::ncclCommWatchdog() + 0x14d (0x790f713ceefd in /usr/local/lib/python3.11/dist-packages/torch/lib/libtorch_cuda.so)
frame #4: <unknown function> + 0xdc253 (0x790f611d8253 in /lib/x86_64-linux-gnu/libstdc++.so.6)
frame #5: <unknown function> + 0x94ac3 (0x790fc7850ac3 in /lib/x86_64-linux-gnu/libc.so.6)
frame #6: clone + 0x44 (0x790fc78e1a04 in /lib/x86_64-linux-gnu/libc.so.6)

terminate called after throwing an instance of 'c10::DistBackendError'
  what():  [PG ID 0 PG GUID 0(default_pg) Rank 2] Process group watchdog thread terminated with exception: [Rank 2] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=2, OpType=ALLREDUCE, NumelIn=1, NumelOut=1, Timeout(ms)=600000) ran for 600048 milliseconds before timing out.
Exception raised from checkTimeout at /pytorch/torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:635 (most recent call first):
frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x790fc69785e8 in /usr/local/lib/python3.11/dist-packages/torch/lib/libc10.so)
frame #1: c10d::ProcessGroupNCCL::WorkNCCL::checkTimeout(std::optional<std::chrono::duration<long, std::ratio<1l, 1000l> > >) + 0x23d (0x790f713cba6d in /usr/local/lib/python3.11/dist-packages/torch/lib/libtorch_cuda.so)
frame #2: c10d::ProcessGroupNCCL::watchdogHandler() + 0xc80 (0x790f713cd7f0 in /usr/local/lib/python3.11/dist-packages/torch/lib/libtorch_cuda.so)
frame #3: c10d::ProcessGroupNCCL::ncclCommWatchdog() + 0x14d (0x790f713ceefd in /usr/local/lib/python3.11/dist-packages/torch/lib/libtorch_cuda.so)
frame #4: <unknown function> + 0xdc253 (0x790f611d8253 in /lib/x86_64-linux-gnu/libstdc++.so.6)
frame #5: <unknown function> + 0x94ac3 (0x790fc7850ac3 in /lib/x86_64-linux-gnu/libc.so.6)
frame #6: clone + 0x44 (0x790fc78e1a04 in /lib/x86_64-linux-gnu/libc.so.6)

Exception raised from ncclCommWatchdog at /pytorch/torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:1905 (most recent call first):
frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x790fc69785e8 in /usr/local/lib/python3.11/dist-packages/torch/lib/libc10.so)
frame #1: <unknown function> + 0x11b4abe (0x790f7139dabe in /usr/local/lib/python3.11/dist-packages/torch/lib/libtorch_cuda.so)
frame #2: <unknown function> + 0xe07bed (0x790f70ff0bed in /usr/local/lib/python3.11/dist-packages/torch/lib/libtorch_cuda.so)
frame #3: <unknown function> + 0xdc253 (0x790f611d8253 in /lib/x86_64-linux-gnu/libstdc++.so.6)
frame #4: <unknown function> + 0x94ac3 (0x790fc7850ac3 in /lib/x86_64-linux-gnu/libc.so.6)
frame #5: clone + 0x44 (0x790fc78e1a04 in /lib/x86_64-linux-gnu/libc.so.6)

[rank6]:[E312 23:52:48.865788709 ProcessGroupNCCL.cpp:684] [Rank 6] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data.
[rank6]:[E312 23:52:48.865815177 ProcessGroupNCCL.cpp:698] [Rank 6] To avoid data inconsistency, we are taking the entire process down.
[rank6]:[E312 23:52:48.867172243 ProcessGroupNCCL.cpp:1899] [PG ID 0 PG GUID 0(default_pg) Rank 6] Process group watchdog thread terminated with exception: [Rank 6] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=2, OpType=ALLREDUCE, NumelIn=1, NumelOut=1, Timeout(ms)=600000) ran for 600082 milliseconds before timing out.
Exception raised from checkTimeout at /pytorch/torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:635 (most recent call first):
frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x7b9ed57785e8 in /usr/local/lib/python3.11/dist-packages/torch/lib/libc10.so)
frame #1: c10d::ProcessGroupNCCL::WorkNCCL::checkTimeout(std::optional<std::chrono::duration<long, std::ratio<1l, 1000l> > >) + 0x23d (0x7b9e805cba6d in /usr/local/lib/python3.11/dist-packages/torch/lib/libtorch_cuda.so)
frame #2: c10d::ProcessGroupNCCL::watchdogHandler() + 0xc80 (0x7b9e805cd7f0 in /usr/local/lib/python3.11/dist-packages/torch/lib/libtorch_cuda.so)
frame #3: c10d::ProcessGroupNCCL::ncclCommWatchdog() + 0x14d (0x7b9e805ceefd in /usr/local/lib/python3.11/dist-packages/torch/lib/libtorch_cuda.so)
frame #4: <unknown function> + 0xdc253 (0x7b9e703d8253 in /lib/x86_64-linux-gnu/libstdc++.so.6)
frame #5: <unknown function> + 0x94ac3 (0x7b9ed690aac3 in /lib/x86_64-linux-gnu/libc.so.6)
frame #6: clone + 0x44 (0x7b9ed699ba04 in /lib/x86_64-linux-gnu/libc.so.6)

terminate called after throwing an instance of 'c10::DistBackendError'
  what():  [PG ID 0 PG GUID 0(default_pg) Rank 6] Process group watchdog thread terminated with exception: [Rank 6] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=2, OpType=ALLREDUCE, NumelIn=1, NumelOut=1, Timeout(ms)=600000) ran for 600082 milliseconds before timing out.
Exception raised from checkTimeout at /pytorch/torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:635 (most recent call first):
frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x7b9ed57785e8 in /usr/local/lib/python3.11/dist-packages/torch/lib/libc10.so)
frame #1: c10d::ProcessGroupNCCL::WorkNCCL::checkTimeout(std::optional<std::chrono::duration<long, std::ratio<1l, 1000l> > >) + 0x23d (0x7b9e805cba6d in /usr/local/lib/python3.11/dist-packages/torch/lib/libtorch_cuda.so)
frame #2: c10d::ProcessGroupNCCL::watchdogHandler() + 0xc80 (0x7b9e805cd7f0 in /usr/local/lib/python3.11/dist-packages/torch/lib/libtorch_cuda.so)
frame #3: c10d::ProcessGroupNCCL::ncclCommWatchdog() + 0x14d (0x7b9e805ceefd in /usr/local/lib/python3.11/dist-packages/torch/lib/libtorch_cuda.so)
frame #4: <unknown function> + 0xdc253 (0x7b9e703d8253 in /lib/x86_64-linux-gnu/libstdc++.so.6)
frame #5: <unknown function> + 0x94ac3 (0x7b9ed690aac3 in /lib/x86_64-linux-gnu/libc.so.6)
frame #6: clone + 0x44 (0x7b9ed699ba04 in /lib/x86_64-linux-gnu/libc.so.6)

Exception raised from ncclCommWatchdog at /pytorch/torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:1905 (most recent call first):
frame #0: c10::Error::Error(c10::SourceLocation, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >) + 0x98 (0x7b9ed57785e8 in /usr/local/lib/python3.11/dist-packages/torch/lib/libc10.so)
frame #1: <unknown function> + 0x11b4abe (0x7b9e8059dabe in /usr/local/lib/python3.11/dist-packages/torch/lib/libtorch_cuda.so)
frame #2: <unknown function> + 0xe07bed (0x7b9e801f0bed in /usr/local/lib/python3.11/dist-packages/torch/lib/libtorch_cuda.so)
frame #3: <unknown function> + 0xdc253 (0x7b9e703d8253 in /lib/x86_64-linux-gnu/libstdc++.so.6)
frame #4: <unknown function> + 0x94ac3 (0x7b9ed690aac3 in /lib/x86_64-linux-gnu/libc.so.6)
frame #5: clone + 0x44 (0x7b9ed699ba04 in /lib/x86_64-linux-gnu/libc.so.6)

W0312 23:52:50.209000 2052 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2120 closing signal SIGTERM
W0312 23:52:50.211000 2052 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2121 closing signal SIGTERM
W0312 23:52:50.213000 2052 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2123 closing signal SIGTERM
W0312 23:52:50.214000 2052 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2124 closing signal SIGTERM
W0312 23:52:50.215000 2052 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2125 closing signal SIGTERM
W0312 23:52:50.216000 2052 torch/distributed/elastic/multiprocessing/api.py:900] Sending process 2127 closing signal SIGTERM
E0312 23:52:52.648000 2052 torch/distributed/elastic/multiprocessing/api.py:874] failed (exitcode: -6) local_rank: 2 (pid: 2122) of binary: /usr/bin/python
Traceback (most recent call last):
  File "/usr/local/bin/accelerate", line 8, in <module>
    sys.exit(main())
             ^^^^^^
  File "/usr/local/lib/python3.11/dist-packages/accelerate/commands/accelerate_cli.py", line 50, in main
    args.func(args)
  File "/usr/local/lib/python3.11/dist-packages/accelerate/commands/launch.py", line 1396, in launch_command
    multi_gpu_launcher(args)
  File "/usr/local/lib/python3.11/dist-packages/accelerate/commands/launch.py", line 1023, in multi_gpu_launcher
    distrib_run.run(args)
  File "/usr/local/lib/python3.11/dist-packages/torch/distributed/run.py", line 883, in run
    elastic_launch(
  File "/usr/local/lib/python3.11/dist-packages/torch/distributed/launcher/api.py", line 139, in __call__
    return launch_agent(self._config, self._entrypoint, list(args))
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/dist-packages/torch/distributed/launcher/api.py", line 270, in launch_agent
    raise ChildFailedError(
torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
=====================================================
/tmp/train_on_hf.py FAILED
-----------------------------------------------------
Failures:
[1]:
  time      : 2026-03-12_23:52:50
  host      : faa36c97b6c7
  rank      : 6 (local_rank: 6)
  exitcode  : -6 (pid: 2126)
  error_file: <N/A>
  traceback : Signal 6 (SIGABRT) received by PID 2126
-----------------------------------------------------
Root Cause (first observed failure):
[0]:
  time      : 2026-03-12_23:52:50
  host      : faa36c97b6c7
  rank      : 2 (local_rank: 2)
  exitcode  : -6 (pid: 2122)
  error_file: <N/A>
  traceback : Signal 6 (SIGABRT) received by PID 2122
=====================================================

>>> Training finished with exit code 1
Ended: 2026-03-12 23:52:53 UTC
>>> Saving training log to HuggingFace...