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

Fix padding_idx #942

Merged
merged 2 commits into from
Mar 10, 2023
Merged

Fix padding_idx #942

merged 2 commits into from
Mar 10, 2023

Conversation

marcoyang1998
Copy link
Collaborator

Someone mentioned in #923 that the non-streaming Transducer models tend to emit the first non-blank symbol at the very beginning. This PR alleviates this issue by removing the padding_idx in the decoder. Here are some results of pruned_transducer_stateless7 on LibriSpeech:

model test-clean test-other average first symbol time (ms)
padding_idx=0 2.19 5.16 7.6
padding_idx=None 2.22 5.14 127

If someone needs this model, I can upload it to huggingface.

@marcoyang1998
Copy link
Collaborator Author

@marcoyang1998 marcoyang1998 merged commit 9ddd811 into k2-fsa:master Mar 10, 2023
@desh2608
Copy link
Collaborator

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants