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

Allow partial for intersect_dense_pruned #1218

Merged
merged 25 commits into from
Sep 23, 2023
Merged

Conversation

pkufool
Copy link
Collaborator

@pkufool pkufool commented Jul 7, 2023

No description provided.

@pkufool pkufool added the ready Ready for review and trigger GitHub actions to run label Jul 19, 2023
@pkufool pkufool removed the ready Ready for review and trigger GitHub actions to run label Aug 6, 2023
@pkufool pkufool added the ready Ready for review and trigger GitHub actions to run label Aug 7, 2023
@pkufool pkufool added ready Ready for review and trigger GitHub actions to run and removed ready Ready for review and trigger GitHub actions to run labels Aug 7, 2023
@pkufool pkufool removed the ready Ready for review and trigger GitHub actions to run label Aug 7, 2023
@pkufool pkufool added the ready Ready for review and trigger GitHub actions to run label Aug 7, 2023
@geoph9
Copy link

geoph9 commented Aug 25, 2023

Is there any plan to add the allow_partial option on intersect and intersect_device? I am doing whole lattice rescoring with a 4-gram LM, and for some reason the lattice for some of the utterances does not have any final states and hence k2.connect(lattice) outputs an empty Fsa.

I have tried increasing the beam size to no avail. I think a workaround would be to allow partial paths, i.e. if not final states are encountered in a lattice then we assume that all states on the last frames are final.

@pkufool
Copy link
Collaborator Author

pkufool commented Aug 26, 2023

Is there any plan to add the allow_partial option on intersect and intersect_device? I am doing whole lattice rescoring with a 4-gram LM, and for some reason the lattice for some of the utterances does not have any final states and hence k2.connect(lattice) outputs an empty Fsa.

I have tried increasing the beam size to no avail. I think a workaround would be to allow partial paths, i.e. if not final states are encountered in a lattice then we assume that all states on the last frames are final.

Yes, I think it is doable, @danpovey what do you think?

@pkufool pkufool added ready Ready for review and trigger GitHub actions to run and removed ready Ready for review and trigger GitHub actions to run labels Sep 8, 2023
Copy link
Collaborator

@danpovey danpovey left a comment

Choose a reason for hiding this comment

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

Thanks!
I am surprised this didn't already have an allow_partial option.

k2/csrc/intersect_dense_pruned.cu Show resolved Hide resolved
@pkufool pkufool added ready Ready for review and trigger GitHub actions to run and removed ready Ready for review and trigger GitHub actions to run labels Sep 22, 2023
@pkufool pkufool added ready Ready for review and trigger GitHub actions to run and removed ready Ready for review and trigger GitHub actions to run labels Sep 22, 2023
@pkufool
Copy link
Collaborator Author

pkufool commented Sep 23, 2023

Merging now, I found the failed tests (mainly windows build) existed for several months, we should fix it in a separate PR.

@pkufool pkufool merged commit 1f11a51 into k2-fsa:master Sep 23, 2023
28 of 144 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
ready Ready for review and trigger GitHub actions to run
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants