Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Upload extracted codebook indexes #429

Merged
merged 17 commits into from
Jun 21, 2022
Merged

Upload extracted codebook indexes #429

merged 17 commits into from
Jun 21, 2022

Conversation

yaozengwei
Copy link
Collaborator

This PR is about pruned_transducer_stateless6. It aims to allow people directly download the extracted codebook indexes for model distillation.

@yaozengwei
Copy link
Collaborator Author

The detailed steps are in distillation_with_hubert.sh.
To directly download the extracted codebook indexes for model distillation, you can set stage=2, stop_stage=4, use_extracted_codebook=True.
To start from scratch, you can set stage=0, stop_stage=4, use_extracted_codebook=False.

@pzelasko
Copy link
Collaborator

I see you fixed the same thing as I in #430; can you also update the HUBERT download URL? Then we could close my PR to avoid conflicts.

@yaozengwei
Copy link
Collaborator Author

I see you fixed the same thing as I in #430; can you also update the HUBERT download URL? Then we could close my PR to avoid conflicts.

Sure. I have also updated the HUBERT download URL. See

wget -c https://dl.fbaipublicfiles.com/hubert/${teacher_model_id}.pt -P ${hubert_model_dir}

@pzelasko
Copy link
Collaborator

Thanks!


# teacher_model_id can be one of
# "hubert_xtralarge_ll60k_finetune_ls960" -> fine-tuned model, it is the one we currently use.
# "hubert_xtralarge_ll60k.pt" -> pretrained model without fintuing
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe should remove the ".pt" suffix of "hubert_xtralarge_ll60k.pt"?

@yaozengwei yaozengwei merged commit d3daeaf into k2-fsa:master Jun 21, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants