Loading dataset...
CachedTrajectoryDataset: 50 demos, 1600 samples
  Source: {'cache_root': '/data/libero/ood_translation_v1', 'benchmark': 'libero_spatial', 'task_ids': [0], 'frame_stride': 3}
  Total: 1600 samples
✓ Train: 1520 samples
✓ Val: 80 samples

Initializing model (type=act)...
Loading ResNet-18 from torchvision (pretrained)...
✓ ResNet-18 backbone is trainable
✓ CLIP projection: 512 → 512
✓ ACT model: 14,093,952 / 14,093,952 trainable params
  Input: CLS(512) + start_kp(2) + eef_pos(3) + gripper(1) = 1030
  pos_mlp:     → (B, 4, 3) [sigmoid, normalized]
  rot_mlp:     → (B, 4, 3) [sigmoid, normalized]
  gripper_mlp: → (B, 4)    [sigmoid, normalized]
Trainable parameters: 14,093,952 / 14,093,952 (100.00%)

Computing dataset stats from random subset: 500/1600 samples (seed=42)
/data2/cameron/miniconda3/envs/uva/lib/python3.10/site-packages/wandb/sdk/data_types/image.py:324: DeprecationWarning: 'mode' parameter is deprecated and will be removed in Pillow 13 (2026-10-15)
✓ Saved stats cache: /data/cameron/567_augmentation_viewpoint_project/checkpoints/act_resnet_multivp_noaug/dataset_stats.json
✓ Height range from dataset: [0.916988, 1.177143] m
✓ Gripper range from dataset: [-1.000000, 1.000000]
✓ Rotation range (delta rotvec): ['-0.016', '-0.098', '-0.006'] .. ['0.051', '0.068', '0.019']
✓ Reference rotation: ['0.9996', '0.0012', '-0.0295', '0.0001']
✓ Position range from dataset: ['-0.408', '-0.336', '0.917'] .. ['0.124', '0.264', '1.177']

Starting training for 1000 epochs...
✓ Rotation loss SKIPPED
✓ EMA loss weights initialized (vol=11.8, rot=3.5, grip=0.69)
  self._image = pil_image.fromarray(
Epochs:   0%|          | 3/1000 [15:50<88:27:53, 319.43s/it]

============================================================
Epoch 0/1000
============================================================
                                                                                                          

  [async eval] failed to log video: [Errno 36] File name too long: '/data/cameron/567_augmentation_viewpoint_project/logs/scratch/eval_run/videos/task_0/ep000_fail.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.h264.mp4'
Train Loss: 0.3622 (Volume: 0.0388, Gripper: 0.3234, Rotation: 0.0000)
Val - Loss: 0.3354, Volume: 0.0112, Pixel Error: 20.12px, Height Error: 0.000mm, Gripper: 0.1062
✓ Saved best model (val_loss=0.3354)

============================================================
Epoch 1/1000
============================================================
Train Loss: 0.0253 (Volume: 0.0118, Gripper: 0.0134, Rotation: 0.0000)
Val - Loss: 0.2923, Volume: 0.0043, Pixel Error: 12.04px, Height Error: 0.000mm, Gripper: 0.0875
✓ Saved best model (val_loss=0.2923)

============================================================
Epoch 2/1000
============================================================

  [async eval] launched at step 500
Train Loss: 0.0151 (Volume: 0.0076, Gripper: 0.0075, Rotation: 0.0000)
Val - Loss: 0.2568, Volume: 0.0024, Pixel Error: 8.58px, Height Error: 0.000mm, Gripper: 0.0688
✓ Saved best model (val_loss=0.2568)

============================================================
Epoch 3/1000
============================================================
Train Loss: 0.0179 (Volume: 0.0090, Gripper: 0.0089, Rotation: 0.0000)
Val - Loss: 0.2454, Volume: 0.0020, Pixel Error: 7.55px, Height Error: 0.000mm, Gripper: 0.0437
✓ Saved best model (val_loss=0.2454)

============================================================
Epoch 4/1000
============================================================
Train Loss: 0.0112 (Volume: 0.0057, Gripper: 0.0055, Rotation: 0.0000)
Val - Loss: 0.2542, Volume: 0.0018, Pixel Error: 7.57px, Height Error: 0.000mm, Gripper: 0.0563

============================================================
Epoch 5/1000
============================================================

  [async eval] launched at step 1000

  Saved step checkpoint: step_1000.pth
Train Loss: 0.0117 (Volume: 0.0059, Gripper: 0.0058, Rotation: 0.0000)
Val - Loss: 0.2427, Volume: 0.0019, Pixel Error: 7.83px, Height Error: 0.000mm, Gripper: 0.0688
✓ Saved best model (val_loss=0.2427)

⏱ Time limit reached (30.5 / 30 min). Stopping.