Loading weights:   0%|          | 0/473 [00:00<?, ?it/s]Loading weights:  63%|██████▎   | 299/473 [00:00<00:00, 2988.41it/s]Loading weights: 100%|██████████| 473/473 [00:00<00:00, 4252.64it/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', 'epoch': 1, 'samples': 100, 'total': 1010, 'steps': 12, 'seconds': 44}
{'phase': 'component_branch', 'epoch': 1, 'samples': 200, 'total': 1010, 'steps': 25, 'seconds': 86}
{'phase': 'component_branch', 'epoch': 1, 'samples': 300, 'total': 1010, 'steps': 37, 'seconds': 129}
{'phase': 'component_branch', 'epoch': 1, 'samples': 400, 'total': 1010, 'steps': 50, 'seconds': 171}
{'phase': 'component_branch', 'epoch': 1, 'samples': 500, 'total': 1010, 'steps': 62, 'seconds': 214}
{'phase': 'component_branch', 'epoch': 1, 'samples': 600, 'total': 1010, 'steps': 75, 'seconds': 256}
{'phase': 'component_branch', 'epoch': 1, 'samples': 700, 'total': 1010, 'steps': 87, 'seconds': 299}
{'phase': 'component_branch', 'epoch': 1, 'samples': 800, 'total': 1010, 'steps': 100, 'seconds': 342}
{'phase': 'component_branch', 'epoch': 1, 'samples': 900, 'total': 1010, 'steps': 112, 'seconds': 384}
{'phase': 'component_branch', 'epoch': 1, 'samples': 1000, 'total': 1010, 'steps': 125, 'seconds': 427}
{'phase': 'component_branch', 'epoch': 2, 'samples': 100, 'total': 1010, 'steps': 139, 'seconds': 475}
{'phase': 'component_branch', 'epoch': 2, 'samples': 200, 'total': 1010, 'steps': 152, 'seconds': 518}
{'phase': 'component_branch', 'epoch': 2, 'samples': 300, 'total': 1010, 'steps': 164, 'seconds': 562}
{'phase': 'component_branch', 'epoch': 2, 'samples': 400, 'total': 1010, 'steps': 177, 'seconds': 605}
{'phase': 'component_branch', 'epoch': 2, 'samples': 500, 'total': 1010, 'steps': 189, 'seconds': 648}
{'phase': 'component_branch', 'epoch': 2, 'samples': 600, 'total': 1010, 'steps': 202, 'seconds': 691}
{'phase': 'component_branch', 'epoch': 2, 'samples': 700, 'total': 1010, 'steps': 214, 'seconds': 733}
{'phase': 'component_branch', 'epoch': 2, 'samples': 800, 'total': 1010, 'steps': 227, 'seconds': 776}
{'phase': 'component_branch', 'epoch': 2, 'samples': 900, 'total': 1010, 'steps': 239, 'seconds': 819}
{'phase': 'component_branch', 'epoch': 2, 'samples': 1000, 'total': 1010, 'steps': 252, 'seconds': 861}
[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.
Loading weights:   0%|          | 0/473 [00:00<?, ?it/s]Loading weights:  34%|███▍      | 162/473 [00:00<00:00, 1617.88it/s]Loading weights: 100%|██████████| 473/473 [00:00<00:00, 3657.97it/s]
{'phase': 'full_pool', 'epoch': 1, 'samples': 100, 'total': 1211, 'steps': 12, 'seconds': 42}
{'phase': 'full_pool', 'epoch': 1, 'samples': 200, 'total': 1211, 'steps': 25, 'seconds': 86}
{'phase': 'full_pool', 'epoch': 1, 'samples': 300, 'total': 1211, 'steps': 37, 'seconds': 129}
{'phase': 'full_pool', 'epoch': 1, 'samples': 400, 'total': 1211, 'steps': 50, 'seconds': 171}
{'phase': 'full_pool', 'epoch': 1, 'samples': 500, 'total': 1211, 'steps': 62, 'seconds': 214}
{'phase': 'full_pool', 'epoch': 1, 'samples': 600, 'total': 1211, 'steps': 75, 'seconds': 257}
{'phase': 'full_pool', 'epoch': 1, 'samples': 700, 'total': 1211, 'steps': 87, 'seconds': 300}
{'phase': 'full_pool', 'epoch': 1, 'samples': 800, 'total': 1211, 'steps': 100, 'seconds': 342}
{'phase': 'full_pool', 'epoch': 1, 'samples': 900, 'total': 1211, 'steps': 112, 'seconds': 385}
{'phase': 'full_pool', 'epoch': 1, 'samples': 1000, 'total': 1211, 'steps': 125, 'seconds': 428}
{'phase': 'full_pool', 'epoch': 1, 'samples': 1100, 'total': 1211, 'steps': 137, 'seconds': 471}
{'phase': 'full_pool', 'epoch': 1, 'samples': 1200, 'total': 1211, 'steps': 150, 'seconds': 513}
{'phase': 'full_pool', 'epoch': 2, 'samples': 100, 'total': 1211, 'steps': 164, 'seconds': 561}
{'phase': 'full_pool', 'epoch': 2, 'samples': 200, 'total': 1211, 'steps': 177, 'seconds': 604}
{'phase': 'full_pool', 'epoch': 2, 'samples': 300, 'total': 1211, 'steps': 189, 'seconds': 647}
{'phase': 'full_pool', 'epoch': 2, 'samples': 400, 'total': 1211, 'steps': 202, 'seconds': 690}
{'phase': 'full_pool', 'epoch': 2, 'samples': 500, 'total': 1211, 'steps': 214, 'seconds': 733}
{'phase': 'full_pool', 'epoch': 2, 'samples': 600, 'total': 1211, 'steps': 227, 'seconds': 775}
{'phase': 'full_pool', 'epoch': 2, 'samples': 700, 'total': 1211, 'steps': 239, 'seconds': 818}
{'phase': 'full_pool', 'epoch': 2, 'samples': 800, 'total': 1211, 'steps': 252, 'seconds': 860}
{'phase': 'full_pool', 'epoch': 2, 'samples': 900, 'total': 1211, 'steps': 264, 'seconds': 903}
{'phase': 'full_pool', 'epoch': 2, 'samples': 1000, 'total': 1211, 'steps': 277, 'seconds': 946}
{'phase': 'full_pool', 'epoch': 2, 'samples': 1100, 'total': 1211, 'steps': 289, 'seconds': 989}
{'phase': 'full_pool', 'epoch': 2, 'samples': 1200, 'total': 1211, 'steps': 302, 'seconds': 1032}
{'state': 'actual_tiny_VLM_LoRA_complete'}
