Add C++ runtime for parakeet-tdt-0.6b-v2.

[csukuangfj]

See also #2180

Usage

1. Build sherpa-onnx
2. Download a model from https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models
   For example

wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8.tar.bz2

tar xvf sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8.tar.bz2

3. Run it

./build/bin/sherpa-onnx-offline \
  --encoder=./sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8/encoder.int8.onnx \
  --decoder=./sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8/decoder.int8.onnx \
  --joiner=./sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8/joiner.int8.onnx \
  --tokens=./sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8/tokens.txt \
  --model-type=nemo_transducer \
  ./sherpa-onnx-nemo-parakeet-tdt-0.6b-v2-int8/test_wavs/0.wav

The output on my mac is

Info about my mac

---

CC @hainan-xv

[csukuangfj]

You can try it by visiting our huggingface space
https://huggingface.co/spaces/k2-fsa/automatic-speech-recognition

(For Chinese users, you can also use the mirror
https://hf.qhduan.com/spaces/k2-fsa/automatic-speech-recognition
)

[csukuangfj]

Here is the RTF test on my orangpi5 max, which uses rk3588.
It has 4 cortex A76 CPUs and 4 cortex A55 CPUs.

RTF on 1 cortex A76 CPU

RTF on 1 cortex A55 CPU

Info about the board

[csukuangfj]

We have provided an Android APK to try this model.

Please see
https://k2-fsa.github.io/sherpa/onnx/vad/apk-asr.html