onnx export

游雁

2023-02-27 f98c4bf6d2bb5202488cd4243efdbca65288c313

 funasr/runtime/python/onnxruntime/README.md

@@ -11,40 +11,38 @@

### Steps:
1. Export the model.
   
   `Tips`: torch 1.11.0 is required.
   - Command: (`Tips`: torch 1.11.0 is required.)

   ```shell
   python -m funasr.export.export_model [model_name] [export_dir] [true]
   ```
   `model_name`: the model is to export.

   `export_dir`: the dir where the onnx is export.

   More details ref to ([export docs](https://github.com/alibaba-damo-academy/FunASR/tree/main/funasr/export))


   - `e.g.`, Export model from modelscope
      ```shell
      python -m funasr.export.export_model 'damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch' "./export" true
      python -m funasr.export.export_model [model_name] [export_dir] [true]
      ```
   - `e.g.`, Export model from local path, the model'name must be `model.pb`.
      ```shell
      python -m funasr.export.export_model '/mnt/workspace/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch' "./export" true
      ```
      `model_name`: the model is to export.

      `export_dir`: the dir where the onnx is export.

       More details ref to ([export docs](https://github.com/alibaba-damo-academy/FunASR/tree/main/funasr/export))

       - `e.g.`, Export model from modelscope
         ```shell
         python -m funasr.export.export_model 'damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch' "./export" true
         ```
       - `e.g.`, Export model from local path, the model'name must be `model.pb`.
         ```shell
         python -m funasr.export.export_model '/mnt/workspace/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch' "./export" true
         ```


2. Install the `rapid_paraformer`
    - Build the rapid_paraformer `whl`
    ```shell
    git clone https://github.com/alibaba/FunASR.git && cd FunASR
    cd funasr/runtime/python/onnxruntime/rapid_paraformer
    python setup.py bdist_wheel
    ```
2. Install the `rapid_paraformer`.
   - Build the rapid_paraformer `whl`
     ```shell
     git clone https://github.com/alibaba/FunASR.git && cd FunASR
     cd funasr/runtime/python/onnxruntime
     python setup.py bdist_wheel
     ```
   - Install the build `whl`
    ```bash
    pip install dist/rapid_paraformer-0.0.1-py3-none-any.whl
    ```
     ```bash
     pip install dist/rapid_paraformer-0.0.1-py3-none-any.whl
     ```

3. Run the demo.
   - Model_dir: the model path, which contains `model.onnx`, `config.yaml`, `am.mvn`.
@@ -52,7 +50,7 @@
   - Output: `List[str]`: recognition result.
   - Example:
        ```python
        from paraformer_onnx import Paraformer
        from rapid_paraformer 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)