add 16bars num_layers 4 ablation: checkpoint + hparams

add 16bars dropout 0.4 ablation: checkpoint + hparams

add 16bars dropout 0.3 ablation: checkpoint + hparams

chore: drop final_ aliases (renamed to best_)

Upload paper/32bars/best_jazz_model_32bars.pth.tar with huggingface_hub

Upload paper/16bars/best_jazz_model_16bars.pth.tar with huggingface_hub

Upload paper/8bars/best_jazz_model_8bars.pth.tar with huggingface_hub

chore: remove top-level {N}bars/ — moved to paper/{N}bars/

Upload paper/32bars/hparams.yaml with huggingface_hub

Upload paper/32bars/final_jazz_model_32bars.pth.tar with huggingface_hub

Upload paper/16bars/hparams.yaml with huggingface_hub

Upload paper/16bars/final_jazz_model_16bars.pth.tar with huggingface_hub

Upload paper/8bars/hparams.yaml with huggingface_hub

Upload paper/8bars/final_jazz_model_8bars.pth.tar with huggingface_hub

Upload 32bars/hparams.yaml with huggingface_hub

Upload 16bars/hparams.yaml with huggingface_hub

Upload 8bars/hparams.yaml with huggingface_hub

Add 32-bar trained model (Phase 2 epoch 50, pitch_acc=0.0988)

Remove old baseline/ path (renamed to 8bars/)

Rename baseline/ → 8bars/ for consistency with 16bars/

Add 16-bar trained model (Phase 2 epoch 70, pitch_acc=0.1773)

Add model

initial commit