README.md

---
dataset_info:
  features:
  - name: audio
    dtype: audio
  - name: text
    dtype: string
configs:
- config_name: default
  data_files:
  - split: train
    path: train/metadata.jsonl
  - split: validation
    path: validation/metadata.jsonl
  - split: test
    path: test/metadata.jsonl
---

# Amicus_AISHELL3

Self-contained Amicus speech dataset packaged for Hugging Face `audiofolder`.

## Splits

- train: 59693 samples, 59.4776 hours
- validation: 3569 samples, 3.6945 hours
- test: 24751 samples, 22.4399 hours

## Load with Hugging Face Datasets

```python
from datasets import load_dataset

ds = load_dataset("audiofolder", data_dir=".")
print(ds)
print(ds["train"][0]["audio"])
```

After uploading this folder to a dataset repository, replace `data_dir="."` with
the dataset repo id if automatic builder detection works for your repository:

```python
from datasets import load_dataset

ds = load_dataset("YOUR_NAMESPACE/Amicus_AISHELL3")
```

## Use with Amicus

Download the whole dataset repository, including Git LFS files, then launch
Amicus training from the downloaded dataset root so relative `audio_path` values
resolve correctly.

```bash
huggingface-cli download YOUR_NAMESPACE/Amicus_AISHELL3 \
  --repo-type dataset \
  --local-dir data/Amicus_AISHELL3

cd data/Amicus_AISHELL3
python /path/to/Amicus/training/stage1/1_semantic_alignment.py \
  --train_data train.jsonl
```

## Source

This dataset is based on AISHELL3, available from OpenSLR:
https://www.openslr.org/93/

## Citation

If you use this dataset, please cite the original AISHELL-3 paper:

```bibtex
@inproceedings{AISHELL-3_2020,
  title={AISHELL-3: A Multi-speaker Mandarin TTS Corpus and the Baselines},
  author={Yao Shi, Hui Bu, Xin Xu, Shaoji Zhang, Ming Li},
  year={2015},
  url={https://arxiv.org/abs/2010.11567}
}
```