[2026-01-08 21:16:30,080][fusion_bench.programs.fusion_program][INFO] - Running the model fusion program.
[2026-01-08 21:16:30,083][fusion_bench.programs.fusion_program][INFO] - loading model pool
[2026-01-08 21:16:32,559][fusion_bench.programs.fusion_program][INFO] - loading method
[2026-01-08 21:16:32,625][fusion_bench.method.classification.image_classification_finetune][INFO] - Training interval: step
[2026-01-08 21:16:32,626][fusion_bench.method.classification.image_classification_finetune][INFO] - Max epochs: -1, max steps: 4000
[2026-01-08 21:16:32,628][fusion_bench.programs.fusion_program][INFO] - loading task pool
[2026-01-08 21:16:34,235][fusion_bench.method.classification.image_classification_finetune][INFO] - Number of classes for dataset kmnist: 10
[2026-01-08 21:16:34,251][datasets.load][WARNING] - Using the latest cached version of the dataset since tanganke/kmnist couldn't be found on the Hugging Face Hub (offline mode is enabled).
[2026-01-08 21:16:34,260][datasets.packaged_modules.cache.cache][WARNING] - Found the latest cached dataset configuration 'kmnist' at /data/dataset/datasets/tanganke___kmnist/kmnist/0.0.0/6b345bef535b98131f1d945d552906452d19b2f9 (last modified on Mon Jan  5 04:58:54 2026).
[2026-01-08 21:16:34,654][fusion_bench.method.classification.image_classification_finetune][INFO] - Training dataset size: 60000
[2026-01-08 21:16:34,687][datasets.load][WARNING] - Using the latest cached version of the dataset since tanganke/kmnist couldn't be found on the Hugging Face Hub (offline mode is enabled).
[2026-01-08 21:16:34,696][datasets.packaged_modules.cache.cache][WARNING] - Found the latest cached dataset configuration 'kmnist' at /data/dataset/datasets/tanganke___kmnist/kmnist/0.0.0/6b345bef535b98131f1d945d552906452d19b2f9 (last modified on Mon Jan  5 04:58:54 2026).
[2026-01-08 21:16:34,766][fusion_bench.method.classification.image_classification_finetune][INFO] - optimizer:
SGD (
Parameter Group 0
    dampening: 0
    differentiable: False
    foreach: None
    fused: None
    lr: 0.01
    maximize: False
    momentum: 0.9
    nesterov: False
    weight_decay: 0.0001
)
[2026-01-08 21:39:25,076][fusion_bench.method.classification.image_classification_finetune][INFO] - Saving the final model to outputs/convnext-base-224/kmnist/batch_size=64,lr=0.01/raw_checkpoints/final