search

sooftware / kospeech

Open-Source Toolkit for End-to-End Korean Automatic Speech Recognition leveraging PyTorch and Hydra.
https://sooftware.github.io/kospeech/
Apache License 2.0
603 stars 191 forks source link

Can you provide some pre-trained model? #136

Closed ZhangDongyuCN closed 3 years ago

ZhangDongyuCN commented 3 years ago

Title

Description

Training this model requires a large number of data sets, which is difficult to train. Can a pre-trained model be provided? I need a Conformer model with LibriSpeech dataset. And, I need this model exprot to onnx, but this can do by myself.

sooftware commented 3 years ago

I have no plan to train LibriSpeech with KoSpeech now. But I'm going to train LibriSpeech using this repo.

ZhangDongyuCN commented 3 years ago

Do you have pre-trained Conformer model with other dataset ? It it still ok for me. Now I just need a pre-trained Conformer model, whatever the dataset.

sooftware commented 3 years ago

I trained conformer in Korean before, but the model code is very different from now, so you won't be able to use it.

ZhangDongyuCN commented 3 years ago

I trained conformer in Korean before, but the model code is very different from now, so you won't be able to use it.

Is the new code in this repo sooftware/conformer? Can you provide the source code and ckpt model, which you used to train Conformer?

sooftware commented 3 years ago

Yes, a little bit different. Unfortunately I have no source code I tranined. You have to train. Sorry.