asimov-groot-n17 / logs /closedloop_mid.log
squaredcuber's picture
logs
5c1609c verified
Raw History Blame Contribute Delete
17.3 kB
GL backend: egl; seeds: checked against 926 dataset seeds: no overlap (max training seed 1200300)
/root/work/Isaac-GR00T/.venv/lib/python3.12/site-packages/albumentations/__init__.py:13: UserWarning: A new version of Albumentations is available: 2.0.8 (you have 1.4.18). Upgrade using: pip install -U albumentations. To disable automatic update checks, set the environment variable NO_ALBUMENTATIONS_UPDATE to 1.
check_for_updates()
Flash Attention 2 only supports torch.float16 and torch.bfloat16 dtypes, but the current dype in Qwen3VLForConditionalGeneration is torch.float32. You should run training or inference using Automatic Mixed-Precision via the `with torch.autocast(device_type='torch_device'):` decorator, or load the model with the `dtype` argument. Example: `model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="flash_attention_2", dtype=torch.float16)`
Flash Attention 2 only supports torch.float16 and torch.bfloat16 dtypes, but the current dype in Qwen3VLModel is torch.float32. You should run training or inference using Automatic Mixed-Precision via the `with torch.autocast(device_type='torch_device'):` decorator, or load the model with the `dtype` argument. Example: `model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="flash_attention_2", dtype=torch.float16)`
Flash Attention 2 only supports torch.float16 and torch.bfloat16 dtypes, but the current dype in Qwen3VLVisionModel is torch.float32. You should run training or inference using Automatic Mixed-Precision via the `with torch.autocast(device_type='torch_device'):` decorator, or load the model with the `dtype` argument. Example: `model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="flash_attention_2", dtype=torch.float16)`
Flash Attention 2 only supports torch.float16 and torch.bfloat16 dtypes, but the current dype in Qwen3VLTextModel is torch.float32. You should run training or inference using Automatic Mixed-Precision via the `with torch.autocast(device_type='torch_device'):` decorator, or load the model with the `dtype` argument. Example: `model = AutoModel.from_pretrained("openai/whisper-tiny", attn_implementation="flash_attention_2", dtype=torch.float16)`
/root/work/Isaac-GR00T/gr00t/model/modules/dit.py:278: FutureWarning: Accessing config attribute `compute_dtype` directly via 'AlternateVLDiT' object attribute is deprecated. Please access 'compute_dtype' over 'AlternateVLDiT's config object instead, e.g. 'unet.config.compute_dtype'.
embedding_dim=self.inner_dim, compute_dtype=self.compute_dtype
/root/work/Isaac-GR00T/gr00t/model/modules/dit.py:309: FutureWarning: Accessing config attribute `output_dim` directly via 'AlternateVLDiT' object attribute is deprecated. Please access 'output_dim' over 'AlternateVLDiT's config object instead, e.g. 'unet.config.output_dim'.
self.proj_out_2 = nn.Linear(self.inner_dim, self.output_dim)
Total number of DiT parameters: 1091722240
Total number of SelfAttentionTransformer parameters: 201433088
Loading checkpoint shards: 0%| | 0/3 [00:00<?, ?it/s] Loading checkpoint shards: 100%|β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆ| 3/3 [00:00<00:00, 24.03it/s] Loading checkpoint shards: 100%|β–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆβ–ˆ| 3/3 [00:00<00:00, 23.99it/s]
[push 1/50] FAIL 'push the yellow block onto the green pad' steps=440 target_not_moved
[push 2/50] OK 'push the yellow block onto the blue pad' steps=136
[push 3/50] FAIL 'push the orange block onto the red pad' steps=440 target_not_moved
[push 4/50] OK 'push the green block onto the blue pad' steps=49
[push 5/50] OK 'push the orange block onto the yellow pad' steps=418
[push 6/50] OK 'push the blue block onto the orange pad' steps=273
[push 7/50] OK 'push the blue block onto the blue pad' steps=56
[push 8/50] OK 'push the yellow block onto the blue pad' steps=106
[push 9/50] FAIL 'push the purple block onto the red pad' steps=440 target_short_of_pad
[push 10/50] FAIL 'push the purple block onto the yellow pad' steps=440 target_short_of_pad
[push 11/50] FAIL 'push the blue block onto the purple pad' steps=440 target_not_moved
[push 12/50] FAIL 'push the red block onto the yellow pad' steps=440 target_on_wrong_pad
[push 13/50] FAIL 'push the green block onto the purple pad' steps=440 target_short_of_pad
[push 14/50] FAIL 'push the yellow block onto the yellow pad' steps=440 target_not_moved
[push 15/50] OK 'push the red block onto the yellow pad' steps=66
[push 16/50] FAIL 'push the red block onto the yellow pad' steps=440 target_not_moved
[push 17/50] OK 'push the blue block onto the red pad' steps=127
[push 18/50] FAIL 'push the red block onto the purple pad' steps=440 target_short_of_pad
[push 19/50] OK 'push the red block onto the orange pad' steps=71
[push 20/50] OK 'push the orange block onto the blue pad' steps=134
[push 21/50] FAIL 'push the red block onto the red pad' steps=440 target_not_moved
[push 22/50] FAIL 'push the green block onto the red pad' steps=440 target_not_moved
[push 23/50] OK 'push the purple block onto the orange pad' steps=132
[push 24/50] OK 'push the orange block onto the red pad' steps=131
[push 25/50] FAIL 'push the orange block onto the blue pad' steps=440 target_not_moved
[push 26/50] OK 'push the blue block onto the green pad' steps=77
[push 27/50] OK 'push the purple block onto the orange pad' steps=101
[push 28/50] OK 'push the purple block onto the orange pad' steps=46
[push 29/50] OK 'push the orange block onto the yellow pad' steps=269
[push 30/50] FAIL 'push the red block onto the red pad' steps=440 target_not_moved
[push 31/50] FAIL 'push the yellow block onto the red pad' steps=440 target_not_moved
[push 32/50] FAIL 'push the orange block onto the purple pad' steps=440 target_short_of_pad
[push 33/50] OK 'push the purple block onto the blue pad' steps=137
[push 34/50] OK 'push the purple block onto the orange pad' steps=114
[push 35/50] FAIL 'push the blue block onto the green pad' steps=440 target_not_moved
[push 36/50] FAIL 'push the green block onto the green pad' steps=440 target_not_moved
[push 37/50] FAIL 'push the yellow block onto the blue pad' steps=440 target_short_of_pad
[push 38/50] FAIL 'push the yellow block onto the green pad' steps=440 target_not_moved
[push 39/50] OK 'push the green block onto the green pad' steps=122
[push 40/50] OK 'push the yellow block onto the green pad' steps=104
[push 41/50] FAIL 'push the purple block onto the orange pad' steps=440 target_short_of_pad
[push 42/50] FAIL 'push the red block onto the purple pad' steps=440 target_short_of_pad
[push 43/50] OK 'push the purple block onto the green pad' steps=61
[push 44/50] FAIL 'push the purple block onto the purple pad' steps=440 target_not_moved
[push 45/50] FAIL 'push the yellow block onto the purple pad' steps=440 target_not_moved
[push 46/50] OK 'push the red block onto the yellow pad' steps=80
[push 47/50] FAIL 'push the purple block onto the red pad' steps=440 target_not_moved
[push 48/50] OK 'push the blue block onto the orange pad' steps=78
[push 49/50] FAIL 'push the blue block onto the yellow pad' steps=440 target_short_of_pad
[push 50/50] FAIL 'push the blue block onto the blue pad' steps=440 target_not_moved
[button 1/50] FAIL 'press the blue button' steps=260 no_press
[button 2/50] FAIL 'press the orange button' steps=260 no_press
[button 3/50] FAIL 'press the purple button' steps=260 no_press
[button 4/50] OK 'press the orange button' steps=64
[button 5/50] FAIL 'press the blue button' steps=260 wrong_button_pressed
[button 6/50] FAIL 'press the purple button' steps=260 no_press
[button 7/50] OK 'press the red button' steps=77
[button 8/50] FAIL 'press the blue button' steps=260 no_press
[button 9/50] FAIL 'press the red button' steps=260 wrong_button_pressed
[button 10/50] FAIL 'press the yellow button' steps=260 partial_press
[button 11/50] FAIL 'press the red button' steps=260 no_press
[button 12/50] OK 'press the purple button' steps=64
[button 13/50] FAIL 'press the red button' steps=260 no_press
[button 14/50] OK 'press the purple button' steps=68
[button 15/50] OK 'press the green button' steps=43
[button 16/50] FAIL 'press the red button' steps=260 no_press
[button 17/50] OK 'press the blue button' steps=42
[button 18/50] FAIL 'press the purple button' steps=260 no_press
[button 19/50] FAIL 'press the yellow button' steps=260 wrong_button_pressed
[button 20/50] FAIL 'press the red button' steps=260 no_press
[button 21/50] OK 'press the red button' steps=45
[button 22/50] FAIL 'press the purple button' steps=260 no_press
[button 23/50] FAIL 'press the yellow button' steps=260 no_press
[button 24/50] FAIL 'press the green button' steps=260 no_press
[button 25/50] OK 'press the blue button' steps=45
[button 26/50] FAIL 'press the yellow button' steps=260 no_press
[button 27/50] OK 'press the yellow button' steps=68
[button 28/50] OK 'press the green button' steps=42
[button 29/50] FAIL 'press the yellow button' steps=260 no_press
[button 30/50] OK 'press the blue button' steps=101
[button 31/50] FAIL 'press the red button' steps=260 no_press
[button 32/50] FAIL 'press the red button' steps=260 no_press
[button 33/50] OK 'press the yellow button' steps=43
[button 34/50] FAIL 'press the purple button' steps=260 no_press
[button 35/50] OK 'press the orange button' steps=67
[button 36/50] OK 'press the red button' steps=51
[button 37/50] OK 'press the red button' steps=55
[button 38/50] OK 'press the purple button' steps=40
[button 39/50] FAIL 'press the purple button' steps=260 no_press
[button 40/50] FAIL 'press the red button' steps=260 no_press
[button 41/50] FAIL 'press the yellow button' steps=260 no_press
[button 42/50] FAIL 'press the orange button' steps=260 no_press
[button 43/50] FAIL 'press the blue button' steps=260 no_press
[button 44/50] FAIL 'press the yellow button' steps=260 no_press
[button 45/50] FAIL 'press the red button' steps=260 no_press
[button 46/50] OK 'press the orange button' steps=44
[button 47/50] FAIL 'press the purple button' steps=260 no_press
[button 48/50] OK 'press the green button' steps=37
[button 49/50] FAIL 'press the blue button' steps=260 no_press
[button 50/50] FAIL 'press the purple button' steps=260 no_press
[box 1/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 2/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 3/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
[box 4/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 5/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 6/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
[box 7/50] OK 'lift the box and put it on the shelf' steps=180
[box 8/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
[box 9/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
[box 10/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 11/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
[box 12/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 13/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 14/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
[box 15/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
[box 16/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 17/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
[box 18/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 19/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
[box 20/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
[box 21/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 22/50] FAIL 'lift the box and put it on the shelf' steps=440 tipped
[box 23/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
[box 24/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 25/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 26/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 27/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
[box 28/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 29/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 30/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 31/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
[box 32/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 33/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 34/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 35/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 36/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 37/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 38/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 39/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 40/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
[box 41/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 42/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 43/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
[box 44/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 45/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
[box 46/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 47/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 48/50] FAIL 'lift the box and put it on the shelf' steps=440 not_lifted
[box 49/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
[box 50/50] FAIL 'lift the box and put it on the shelf' steps=440 dropped_or_not_placed
# Closed-loop evaluation: groot (/root/work/runs/asimov_n17/checkpoint-6000)
50 held-out episodes per task (seeds 2**40 + task_id * 2**20 + k; checked against 926 dataset seeds: no overlap (max training seed 1200300)), 8 of 16 actions executed per inference at 20 Hz, success held 10 steps, GL egl.
| Task | Success | 95% CI | Mean time to success | Failures |
| --- | --- | --- | --- | --- |
| push | 23/50 (46%) | 0.33-0.60 | 5.83 s | target_not_moved 17, target_short_of_pad 9, target_on_wrong_pad 1 |
| button | 18/50 (36%) | 0.24-0.50 | 2.32 s | no_press 28, wrong_button_pressed 3, partial_press 1 |
| box | 1/50 (2%) | 0.00-0.10 | 8.55 s | not_lifted 30, dropped_or_not_placed 18, tipped 1 |
Policy calls: {'mean': 0.0944, 'p50': 0.0941, 'p90': 0.0954, 'calls': 5762}
## push: per instruction
| Instruction | Success |
| --- | --- |
| push the blue block onto the blue pad | 1/2 |
| push the blue block onto the green pad | 1/2 |
| push the blue block onto the orange pad | 2/2 |
| push the blue block onto the purple pad | 0/1 |
| push the blue block onto the red pad | 1/1 |
| push the blue block onto the yellow pad | 0/1 |
| push the green block onto the blue pad | 1/1 |
| push the green block onto the green pad | 1/2 |
| push the green block onto the purple pad | 0/1 |
| push the green block onto the red pad | 0/1 |
| push the orange block onto the blue pad | 1/2 |
| push the orange block onto the purple pad | 0/1 |
| push the orange block onto the red pad | 1/2 |
| push the orange block onto the yellow pad | 2/2 |
| push the purple block onto the blue pad | 1/1 |
| push the purple block onto the green pad | 1/1 |
| push the purple block onto the orange pad | 4/5 |
| push the purple block onto the purple pad | 0/1 |
| push the purple block onto the red pad | 0/2 |
| push the purple block onto the yellow pad | 0/1 |
| push the red block onto the orange pad | 1/1 |
| push the red block onto the purple pad | 0/2 |
| push the red block onto the red pad | 0/2 |
| push the red block onto the yellow pad | 2/4 |
| push the yellow block onto the blue pad | 2/3 |
| push the yellow block onto the green pad | 1/3 |
| push the yellow block onto the purple pad | 0/1 |
| push the yellow block onto the red pad | 0/1 |
| push the yellow block onto the yellow pad | 0/1 |
Videos: videos/push_failure_seed1099511627776.mp4, videos/push_success_seed1099511627777.mp4
## button: per instruction
| Instruction | Success |
| --- | --- |
| press the blue button | 3/8 |
| press the green button | 3/4 |
| press the orange button | 3/5 |
| press the purple button | 3/11 |
| press the red button | 4/13 |
| press the yellow button | 2/9 |
Videos: videos/button_failure_seed1099512676352.mp4, videos/button_success_seed1099512676355.mp4
## box: per instruction
| Instruction | Success |
| --- | --- |
| lift the box and put it on the shelf | 1/50 |
Videos: videos/box_failure_seed1099513724928.mp4, videos/box_success_seed1099513724934.mp4