Create README.md

# SSL-FT-PRON: Fine-tuned SSL Models for Automatic Pronunciation Assessment (APA)

**Author:** Haeyoung Lee (haeylee)
**Paper:** *Analysis of Various Self-Supervised Learning Models for Automatic Pronunciation Assessment (APSIPA ASC 2024)*
**Code:** https://github.com/hy310/ssl_finetuning

This repository on the Hub is a **collection of sub-checkpoints** for different SSL backbones (Wav2Vec2.0, HuBERT, WavLM) under three training strategies:

- **CTC**: ASR-style head with Connectionist Temporal Classification
- **Freeze**: feature extractor (CNN frontend) frozen during fine-tuning
- **General**: no CTC head; a small regression head predicts four APA scores
*(Accuracy, Fluency, Prosody, Total)*

> Each variant lives in a **subdirectory**. Load it using the full path
> (e.g., `haeylee/ssl_ft_pron/wav2vec2/general/02_wav2vec2-large-960h`).

---

README.md

@@ -0,0 +1,21 @@
+---
+base_model:
+- facebook/wav2vec2-large
+- facebook/wav2vec2-large-960h
+- facebook/wav2vec2-large-lv60
+- facebook/wav2vec2-large-xlsr-53
+- facebook/wav2vec2-xls-r-300m
+- facebook/hubert-large-ll60k
+- facebook/hubert-base-ls960
+- facebook/hubert-xlarge-ll60k
+- facebook/hubert-xlarge-ls960-ft
+- microsoft/wavlm-large
+- microsoft/wavlm-base-plus
+- microsoft/wavlm-base-plus-sv
+tags:
+- self-supervised-learning
+- pronunciation-assessment
+- speech
+metrics:
+- pearsonr
+---