FireRedASR-AED / README.md
inoryQwQ's picture
Trim default Python dependencies
8f43425
|
Raw
History Blame Contribute Delete
1.63 kB
---
license: apache-2.0
---
# FireRedASR-AED
小红书ASR AED-L版本在AX650N上的部署,原项目地址为:[https://github.com/FireRedTeam/FireRedASR](https://github.com/FireRedTeam/FireRedASR)
转换后的模型放置在axmodel目录,目前支持中文、英文,最长输入10秒的音频,超过10秒的音频会用VAD切割后推理。
## 模型转换
[参考Github](https://github.com/ml-inory/FireRedASR.axera/tree/main)
## 支持平台
- [x] AX650N
## 安装依赖
### Audio backend
```
sudo apt install libsndfile1
```
### Python
测试环境为Python 3.12,建议使用[Miniconda](https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-aarch64.sh
),安装方法[参考](https://www.anaconda.com/docs/getting-started/miniconda/install#aws-graviton2%2Farm64)
```
conda create -n fireredasr python=3.12
conda activate fireredasr
pip install -r requirements.txt
```
`requirements.txt` 只包含 AX650N AED 推理路径需要的运行时依赖。其他脚本按需安装:
- `fireredasr_onnx.py`: `pip install onnxruntime`
- `fireredasr/speech2text.py` 的 LLM 路径: `pip install transformers peft`
- `fireredasr/utils/wer.py --do_tn 1`: `pip install cn2an`
### 手动安装pyaxengine
```
wget https://github.com/AXERA-TECH/pyaxengine/releases/download/0.1.3.rc2/axengine-0.1.3-py3-none-any.whl
pip install axengine-0.1.3-py3-none-any.whl
```
## 使用
```
conda activate fireredasr
python test_ax_model.py
```
```hypo_axmodel.txt```包含识别结果
## 性能表现
RTF ~= 0.3
CER(on custom dataset): 3.45%