Add sherpa-onnx support #50
Conversation
|
新增加了TTS引擎部分，现在可以使用 sherpa-onnx TTS了，在config_alts里添加了部分配置样例，方便参考。
推荐一下模型： |
|
太牛了👍 我来看看 |
|
我也在看这个 sherpa-onnx，感觉真不错。正想着他的 tts 部分能替代 Melo 你就更新了，太赞了👍。 |
|
已经更新了 README.md，添加了简要的安装使用说明， 虽然觉得sherpa-onnx也可以代替 funasr 成为默认的推荐，但还是等等后面大家使用的反馈吧。 |
|
抱歉，我这几天期末考有点忙来着。我一两天之后再来看... |
没关系的，还请先好好准备考试。我能理解是责任感在作怪，但真的无需感到压力哦。 |
| # whisper_encoder: "" # Path to the Whisper encoder model (e.g., "path/to/encoder.onnx") | ||
| # whisper_decoder: "" # Path to the Whisper decoder model (e.g., "path/to/decoder.onnx") | ||
| # --- For model_type: "sense_voice" --- | ||
| sense_voice: "/path/to/sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/model.onnx" # Path to the SenseVoice model (e.g., "path/to/model.onnx") |
There was a problem hiding this comment.
I suggest that you use model.int8.onnx, which is way smaller in file size than that of model.onnx.
There was a problem hiding this comment.
Thanks for the heads up! I appreciate the suggestion to use model.int8.onnx. Do you have any insights into how the int8 model performs compared to the original model.onnx in terms of recognition accuracy? Have you had a chance to evaluate its performance?
There was a problem hiding this comment.
I only tested it on fewer than 4 test wave files and the quantized model produces identical results as the not quantized model.
This pull request adds support for sherpa-onnx (https://github.com/k2-fsa/sherpa-onnx) ASR models. It allows users to easily integrate various sherpa-onnx models, including transducer, Paraformer, NeMo CTC, WeNet CTC, Whisper, TDNN CTC, and SenseVoice models.
Note: Only SenseVoice and Paraformer models have been tested at this time. Further testing with other model types is encouraged.
(This implementation was developed with the assistance of an AI language model.)
sherpa-onnx offers great performance and is significantly lighter than FunASR. (no torch)