Upload folder using huggingface_hub

PredANNpp_SongAcc_Acoustic_Pretrain10kEpoch_Finetune3.5kEpoch_seed42.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7eab2d06d24b62044a368525ae6cd168e99559cf405d064e407079bc632952a9
+size 78942947

README.md

@@ -1,3 +1,65 @@
----
-license: cc-by-nc-4.0
----
+---
+language: en
+license: cc-by-nc-4.0
+tags:
+- eeg
+- music
+- representation-learning
+- pytorch-lightning
+- transformer
+---
+
+# PredANN++ (Acoustic, MuQ) — Encoder-only checkpoint
+
+## Model description
+
+This model is a finetuned **EEG encoder** for **song identification from 3-second EEG segments** (10-class classification) on the NMED-T dataset.
+
+The encoder is trained with multitask pretraining (masked prediction) using **MuQ acoustic embeddings**, followed by finetuning with cross-entropy on song ID.
+
+- Input: EEG (128 channels, 125 Hz, 3 seconds)
+- Output: Song ID logits (10 classes)
+
+## Intended use
+
+- Research use for EEG-based music recognition
+- Comparing the effect of acoustic neural representations (MuQ) with expectation-related representations
+
+## Not intended use
+
+- Medical diagnosis
+- Any clinical decision making
+- Commercial usage without verifying upstream non-commercial licenses
+
+## Training data
+
+- NMED-T (Naturalistic Music EEG Dataset – Tempo), 10 songs, 20 subjects, trial=1.
+
+## Training procedure (high-level)
+
+1. Multitask pretraining: encoder-decoder masked prediction of MuQ acoustic tokens (50% masking)
+2. Finetuning: encoder-only training for Song ID classification
+
+## Evaluation
+
+The repository contains an evaluation script:
+- `codes_3s/analysis/evaluate.py`
+
+## License and upstream dependencies (IMPORTANT)
+
+This checkpoint is trained using features computed with:
+- MuQ model weights (CC-BY-NC 4.0)
+
+Therefore, this checkpoint is distributed as **CC-BY-NC 4.0**.
+
+## Citation
+
+If you use this model, please cite:
+- The PredANN++ paper (to be added upon publication)
+- NMED-T dataset paper
+- MuQ model
+
+## Acknowledgements
+
+We borrow and adapt code from multiple repositories. Please see:
+- `THIRD_PARTY_NOTICES.md` in the GitHub repository.