Merge pull request #380 from alibaba-damo-academy/main

yhliang

2023-04-18 9817785c66a13caa681a8e9e272f2ae949233542

 funasr/runtime/python/libtorch/README.md

@@ -1,57 +1,54 @@
## Using funasr with libtorch
# Libtorch-python

[FunASR](https://github.com/alibaba-damo-academy/FunASR) hopes to build a bridge between academic research and industrial applications on speech recognition. By supporting the training & finetuning of the industrial-grade speech recognition model released on ModelScope, researchers and developers can conduct research and production of speech recognition models more conveniently, and promote the development of speech recognition ecology. ASR for Fun！
## Export the model
### Install [modelscope and funasr](https://github.com/alibaba-damo-academy/FunASR#installation)

### Introduction
- Model comes from [speech_paraformer](https://www.modelscope.cn/models/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch/summary).
```shell
pip3 install torch torchaudio
pip install -U modelscope
pip install -U funasr
```

### Steps:
1. Export the model.
   - Command: (`Tips`: torch >= 1.11.0 is required.)
### Export [onnx model](https://github.com/alibaba-damo-academy/FunASR/tree/main/funasr/export)

       More details ref to ([export docs](https://github.com/alibaba-damo-academy/FunASR/tree/main/funasr/export))
```shell
python -m funasr.export.export_model --model-name damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch --export-dir ./export --type torch --quantize True
```

       - `e.g.`, Export model from modelscope
         ```shell
         python -m funasr.export.export_model --model-name damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch --export-dir ./export --type torch --quantize False
         ```
       - `e.g.`, Export model from local path, the model'name must be `model.pb`.
         ```shell
         python -m funasr.export.export_model --model-name ./damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch --export-dir ./export --type torch --quantize False
         ```


2. Install the `funasr_torch`.
## Install the `funasr_torch`.
    
    install from pip
    ```shell
    pip install --upgrade funasr_torch -i https://pypi.Python.org/simple
    ```
    or install from source code
install from pip
```shell
pip install -U funasr_torch
# For the users in China, you could install with the command:
# pip install -U funasr_torch -i https://mirror.sjtu.edu.cn/pypi/web/simple
```
or install from source code

    ```shell
    git clone https://github.com/alibaba/FunASR.git && cd FunASR
    cd funasr/runtime/python/libtorch
    python setup.py build
    python setup.py install
    ```
```shell
git clone https://github.com/alibaba/FunASR.git && cd FunASR
cd funasr/runtime/python/libtorch
pip install -e ./
# For the users in China, you could install with the command:
# pip install -e ./ -i https://mirror.sjtu.edu.cn/pypi/web/simple
```

3. Run the demo.
   - Model_dir: the model path, which contains `model.torchscripts`, `config.yaml`, `am.mvn`.
   - Input: wav formt file, support formats: `str, np.ndarray, List[str]`
   - Output: `List[str]`: recognition result.
   - Example:
        ```python
        from funasr_torch import Paraformer
## Run the demo.
- Model_dir: the model path, which contains `model.torchscripts`, `config.yaml`, `am.mvn`.
- Input: wav formt file, support formats: `str, np.ndarray, List[str]`
- Output: `List[str]`: recognition result.
- Example:
     ```python
     from funasr_torch import Paraformer

        model_dir = "/nfs/zhifu.gzf/export/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch"
        model = Paraformer(model_dir, batch_size=1)
     model_dir = "/nfs/zhifu.gzf/export/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch"
     model = Paraformer(model_dir, batch_size=1)

        wav_path = ['/nfs/zhifu.gzf/export/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch/example/asr_example.wav']
     wav_path = ['/nfs/zhifu.gzf/export/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch/example/asr_example.wav']

        result = model(wav_path)
        print(result)
        ```
     result = model(wav_path)
     print(result)
     ```

## Performance benchmark