Are the pre-trained models available somewhere? about wenet HOT 8 CLOSED

Thanks,
Here https://github.com/mobvoi/wenet/blob/main/examples/librispeech/s0/README.md?
I don't see anything there.

from wenet.

Thanks a lot for providing this! I see that in the pretrained model archives, the bpe models aren't available. If one wants to try these models on different audio files, we would need this file. Otherwise, we have to download all the training data and rebuild to bpe model.

from wenet.

Hey @jprobichaud - I was stuck on the same thing. After I looked a bit more, I was able to find a file (for instance in 20210215_unified_conformer_exp.tar.gz) called "train_960_unigram5000.model" which is the BPE model. This should work for you!

from wenet.

It's on developing now, will provide it once it is ready.

from wenet.

we have uploaded some pre-trained models, the details can be seen in s0/README

from wenet.

There are pre-trained models for AIShell in:
https://github.com/mobvoi/wenet/tree/main/examples/aishell/s0/README.md

Librispeech models need to be uploaded. @whiteshirt0429

from wenet.

I have uploaded a librispeech u2 model，the model link is "http://mobvoi-speech-public.ufile.ucloud.cn/public/wenet/librispeech/20210122_unified_conformer_exp.tar.gz" and you can use git hash "1ea7515e4722f4e05a5985f587120c994a93694a" to load it.

from wenet.

It seems oss is not accessible from outside China. Maybe put in some other places, such as dropbox?

from wenet.