| { |
| "model_type": "clarity-omr", |
| "architectures": [ |
| "DaViT_RoPE_Decoder" |
| ], |
| "pipeline": { |
| "stage_a": { |
| "model": "yolo.pt", |
| "architecture": "YOLOv8m", |
| "task": "staff_detection" |
| }, |
| "stage_b": { |
| "model": "model.safetensors", |
| "architecture": "DaViT encoder + custom RoPE decoder", |
| "encoder": "davit_base.msft_in1k", |
| "encoder_dim": 768, |
| "decoder_dim": 768, |
| "decoder_layers": 8, |
| "decoder_heads": 12, |
| "max_decode_length": 512, |
| "vocab_size": 487, |
| "adaptation": "DoRA", |
| "dora_rank": 64, |
| "task": "staff_recognition" |
| } |
| } |
| } |
