Skip to content

jit: support nn.Moudle.named_parameters/buffers #672

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

Merged
merged 6 commits into from
Jun 28, 2024

Conversation

kshitij12345
Copy link
Collaborator

@kshitij12345 kshitij12345 commented Jun 27, 2024

Fixes: #601

As elaborated in the comment, we add lookaside for Module.named_{parameters/buffers} (for now).

@kshitij12345 kshitij12345 marked this pull request as ready for review June 27, 2024 14:17
Copy link
Collaborator

@t-vi t-vi left a comment

Choose a reason for hiding this comment

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

@kshitij12345 Absolutely fantastic to work on this!
I think we can solve the duplication by first iterating in the lookaside (directly in Python without proxying / interpreter), keeping the names we get there, and then in the _impl filter...

@tfogal
Copy link
Collaborator

tfogal commented Jun 27, 2024

Just wanted to comment that I tried this with NeVA (which is where #601 came from) and the run gets by where we were stopped before! 🎉
We hit another bug before the model actually runs, so this isn't an e2e signal--but a very promising start!

Copy link
Collaborator

@t-vi t-vi left a comment

Choose a reason for hiding this comment

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

Thank you @kshitij12345

@t-vi t-vi merged commit 27a573d into Lightning-AI:main Jun 28, 2024
36 checks passed
@kshitij12345 kshitij12345 deleted the named-params-and-buffers-jit branch June 28, 2024 09:12
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.

unhashable type: 'TensorProxy' error on NeVA model
3 participants