Export to ncnn

We support exporting the following models to ncnn:

• Zipformer transducer models

• LSTM transducer models

• ConvEmformer transducer models

We also provide sherpa-ncnn for performing speech recognition using ncnn with exported models. It has been tested on the following platforms:

• Linux

• macOS

• Windows

• Android

• iOS

• Raspberry Pi

• 爱芯派 (MAIX-III AXera-Pi).

• RV1126

sherpa-ncnn is self-contained and can be statically linked to produce a binary containing everything needed. Please refer to its documentation for details:

• https://k2-fsa.github.io/sherpa/ncnn/index.html

• Export streaming Zipformer transducer models to ncnn
  • 1. Download the pre-trained model
  • 2. Install ncnn and pnnx
  • 3. Export the model via torch.jit.trace()
  • 4. Export torchscript model via pnnx
  • 5. Test the exported models in icefall
  • 6. Modify the exported encoder for sherpa-ncnn
• Export ConvEmformer transducer models to ncnn
  • 1. Download the pre-trained model
  • 2. Install ncnn and pnnx
  • 3. Export the model via torch.jit.trace()
  • 4. Export torchscript model via pnnx
  • 5. Test the exported models in icefall
  • 6. Modify the exported encoder for sherpa-ncnn
  • 7. (Optional) int8 quantization with sherpa-ncnn
• Export LSTM transducer models to ncnn
  • 1. Download the pre-trained model
  • 2. Install ncnn and pnnx
  • 3. Export the model via torch.jit.trace()
  • 4. Export torchscript model via pnnx
  • 5. Test the exported models in icefall
  • 6. Modify the exported encoder for sherpa-ncnn
  • 7. (Optional) int8 quantization with sherpa-ncnn