Loading weights:   0%|          | 0/473 [00:00<?, ?it/s]Loading weights:  37%|███▋      | 174/473 [00:00<00:00, 1739.66it/s]Loading weights: 100%|██████████| 473/473 [00:00<00:00, 3836.95it/s]
[transformers] `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
[transformers] `causal_conv1d_fn` is falling back to its reference PyTorch implementation because `causal_conv1d` is not installed. This is correct but much slower; install `causal_conv1d` for the optimized kernel.
[transformers] `chunk_gated_delta_rule` is falling back to its reference PyTorch implementation because `flash-linear-attention` is not installed. This is correct but much slower; install `flash-linear-attention` for the optimized kernel.
{'phase': 'component_branch', 'format_repair_samples': 200, 'total': 1010, 'steps': 25, 'seconds': 86}
{'phase': 'component_branch', 'format_repair_samples': 400, 'total': 1010, 'steps': 50, 'seconds': 172}
{'phase': 'component_branch', 'format_repair_samples': 600, 'total': 1010, 'steps': 75, 'seconds': 258}
{'phase': 'component_branch', 'format_repair_samples': 800, 'total': 1010, 'steps': 100, 'seconds': 344}
{'phase': 'component_branch', 'format_repair_samples': 1000, 'total': 1010, 'steps': 125, 'seconds': 429}
[transformers] `causal_conv1d_update` is falling back to its reference PyTorch implementation because `causal_conv1d` is not installed. This is correct but much slower; install `causal_conv1d` for the optimized kernel.
[transformers] `fused_recurrent_gated_delta_rule` is falling back to its reference PyTorch implementation because `flash-linear-attention` is not installed. This is correct but much slower; install `flash-linear-attention` for the optimized kernel.
{'phase': 'format_repair_holdout', 'targets': 201, 'JSON_valid': 0, 'attribute_accuracy': {'支腿': 0.0, '横梁': 0.0, '材质': 0.0}, 'latency_ms_mean': 154.1768810400492, 'peak_allocated_GiB': 2.3874387741088867, 'observation_generalization_certified': False, 'observation_labeled_examples': 1}
Loading weights:   0%|          | 0/473 [00:00<?, ?it/s]Loading weights:  53%|█████▎    | 253/473 [00:00<00:00, 2527.40it/s]Loading weights: 100%|██████████| 473/473 [00:00<00:00, 3839.06it/s]
{'phase': 'full_pool', 'format_repair_samples': 200, 'total': 1211, 'steps': 25, 'seconds': 85}
{'phase': 'full_pool', 'format_repair_samples': 400, 'total': 1211, 'steps': 50, 'seconds': 172}
{'phase': 'full_pool', 'format_repair_samples': 600, 'total': 1211, 'steps': 75, 'seconds': 257}
{'phase': 'full_pool', 'format_repair_samples': 800, 'total': 1211, 'steps': 100, 'seconds': 342}
{'phase': 'full_pool', 'format_repair_samples': 1000, 'total': 1211, 'steps': 125, 'seconds': 427}
{'phase': 'full_pool', 'format_repair_samples': 1200, 'total': 1211, 'steps': 150, 'seconds': 513}
{'state': 'actual_tiny_VLM_format_repair_complete'}
