导出模型

1. 命令行导出

funasr-export ++model=paraformer ++quantize=false

2. Python 导出

from funasr import AutoModel

model = AutoModel(model="paraformer")
res = model.export(quantize=False)

3. 测试 ONNX

需要先安装好 funasr-onnx 库。

from funasr_onnx import Paraformer

model_dir = "damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch"
model = Paraformer(model_dir, batch_size=1, quantize=True)

wav_path = [
    '~/.cache/modelscope/hub/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch/example/asr_example.wav'
    ]

result = model(wav_path)
print(result)