Masao Someki

Hi @jinggaizi , you can set `max_seq_len` operation as the following: ```python from espnet_onnx.export import ASRModelExport m = ASRModelExport() m.set_export_config( max_seq_len=1024 ) m.export_from_pretrained(...) ```

Hi @anirpipi, sorry for the late reply, and thank you for reporting the issue. It may be a bug, so I would like to check this problem. It seems you...

Hi @b-flo, thank you! Am I right that the new version of the transducer is implemented in `espnet2/asr_transducer`? Also, I will try to support streaming inference with transducer models with...

Hi @unparalleled-ysj , I have never tried to export the istft operation. It seems that there are some discussion about supporting istft operation, so maybe we have to wait for...

Hi @yuekaizhang, thank you for reporting the issue! I recently found that the current beam search implementation of espnet_onnx might cause a large accuracy loss, and this case might be...

@yuekaizhang @pengaoao I'm sorry for the late replay, I fixed the beam_search-related bug. Would you check the accuracy with the latest espnet_onnx? Note that `ctc_weight=0.3` in the model, so please...

@pengaoao Thank you for reporting, I investigated bugs and found that there is a parity issue with TransformerDecoder. I fixed it in #52, and got the same result sentence with...

Thank you @pengaoao, we have fixed bugs in the stft frontend and now we can get exactly the same result. @yuekaizhang, would you check your performance with the latest version?

@sanjuktasr We can check if we have parity issues with espnet_onnx by running the `test_inference_asr` test, and the contextual conformer block does not have a parity issue.

@PhenixCFLi Thank you for reporting the issue! I will check the details of this problem. Just for clarification, would you tell me the version of the following modules: - onnx...