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.

Sign up for GitHub

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

Jump to bottom

update weights optimization #1791

Open
wants to merge 8 commits into
base: develop
Choose a base branch
from
Open

update weights optimization #1791

wants to merge 8 commits into from
+44 −38

Conversation

clement-pages
Copy link
Collaborator

BREAKING: remove finetune_wavlm attributes from the PixIT task, and replace it with walm_frozen in Totatonet separation model, to be consistent with what is done for the segmentation model.

@clement-pages
Copy link
Collaborator Author

Merging this PR should fix #1789

@hbredin
Copy link
Member

hbredin commented Nov 20, 2024

@Lebourdais suggested that we set wavlm.requires_grad to False in __init__ instead of using the with torch.no_grad logic. I think it is indeed a better idea as it allows to get a better estimate of the number of trainable parameters (i.e. those for which requires_grad is set to True.

@clement-pages can you update your PR to make this change and also update the SSeRiouSS model and diarization task accordingly?

@clement-pages
Copy link
Collaborator Author

@clement-pages can you update your PR to make this change and also update the SSeRiouSS model and diarization task accordingly?

OK, I will update it.

clement-pages added 2 commits November 20, 2024 13:29
use param.requires_grad instead of torch.no_grad()
ca2a5d4
allow to use more than two optimizers
c3a0313 
In the same way that is done for the speaker diarization task
@clement-pages clement-pages changed the title fix weights update issue when training separation model update weights optimization Nov 20, 2024
hbredin
hbredin requested changes Nov 25, 2024
View reviewed changes
pyannote/audio/tasks/segmentation/speaker_diarization.py Outdated Show resolved Hide resolved
pyannote/audio/tasks/separation/PixIT.py Outdated
Comment on lines 1015 to 1017
optimizers = (
[optimizers] if not isinstance(optimizers, list) else optimizers
)
Copy link
Member

Choose a reason for hiding this comment

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

Suggested change
optimizers = (
[optimizers] if not isinstance(optimizers, list) else optimizers
)
optimizers = optimizers if isinstance(optimizers, list) else [optimizers]

hbredin
hbredin reviewed Nov 25, 2024
View reviewed changes
pyannote/audio/models/separation/ToTaToNet.py
@@ -137,6 +139,7 @@ def __init__(
task: Optional[Task] = None,
n_sources: int = 3,
use_wavlm: bool = True,
wavlm_frozen: bool = False,
Copy link
Member

Choose a reason for hiding this comment

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

Did you double check that it does not break loading https://huggingface.co/pyannote/separation-ami-1.0?

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

No, I will check it tomorrow.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@hbredin hbredin hbredin requested changes

Requested changes must be addressed to merge this pull request.

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@clement-pages @hbredin