Binbin Zhang
Binbin Zhang
You can refer https://github.com/kaldi-asr/kaldi/tree/master/egs/swbd/s5 to figure out the problem.
@pengzhendong would you please help on the problem?
please try download the data from modelscope, please see https://github.com/wenet-e2e/WenetSpeech#from-modelscope
看 LOG 提示不是从 modelscope 下载的。
Fixed, see https://github.com/wenet-e2e/wenet/pull/2798. please pull the latest code and try again.
You can just modify the training config`train.yaml`, setting `use_dynamic_chunk` to true.
Yes, that's all. CTC is used for streaming in wenet, and autogressive decoding is used for non-streaming result.
Please see https://github.com/wenet-e2e/wenet/blob/main/wenet/models/whisper/whisper.py