Upload variants/maestro-prompted/README.md with huggingface_hub

variants/maestro-prompted/README.md

@@ -0,0 +1,49 @@
+---
+license: mit
+library_name: pytorch
+tags:
+  - music
+  - audio
+  - midi
+  - error-detection
+  - laddersym
+---
+
+# LadderSym MAESTRO-E (Prompted)
+
+LadderSym checkpoint for MAESTRO-E evaluation with score prompting enabled.
+
+## Variant
+
+- Dataset: MAESTRO-E
+- Prompt setting: prompted (`use_prompt=True`)
+- Run ID: `e4iv21me`
+- Source checkpoint:
+  - `/home/chou150/code/Muse/outputs/2025-01-03/polytunev2_best/interttm_MT3Net_MAESTRO/e4iv21me/checkpoints/last.ckpt`
+
+## Reproduce Evaluation
+
+```bash
+python test_camae_mt3.py \
+  --config-dir=config \
+  --config-name=config_maestro_f1_0.95_prompt \
+  model=laddersym_MT3Net \
+  dataset_is_randomize_tokens=False \
+  path=/path/to/model.ckpt \
+  eval.eval_dataset=MAESTRO \
+  dataset=MAESTRO \
+  eval.exp_tag_name=laddersym_maestro_prompted \
+  hydra/job_logging=disabled \
+  eval.contiguous_inference=True \
+  eval.batch_size=16 \
+  split_frame_length=2000
+```
+
+## Notes
+
+- This repo stores raw PyTorch Lightning checkpoints (`.ckpt`).
+- Inference/evaluation currently uses the LadderSym codebase scripts.
+
+## Citation
+
+Please add the final LadderSym BibTeX entry from the camera-ready paper.