Ensure autograd graphs are freed between attribute calls #1347
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Summary: Some attributions returned by gradient-based methods still have a `grad_fn` from autograd (e.g. `LayerGradientXActivation`). This diff ensures that the autograd graph is freed between attribute calls within `LLMGradientAttribution` to eliminate this as a potential source of VRAM accumulation. Also wrapped `model.generate` with a `no_grad` context to avoid unecessary memory usage. Differential Revision: D62671994
|
This pull request was exported from Phabricator. Differential Revision: D62671994 |
|
This pull request has been merged in b7ca840. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary:
Some attributions returned by gradient-based methods still have a
grad_fnfrom autograd (e.g.LayerGradientXActivation). This diff ensures that the autograd graph is freed between attribute calls withinLLMGradientAttributionto eliminate this as a potential source of VRAM accumulation.Also wrapped
model.generatewith ano_gradcontext to avoid unecessary memory usage.Differential Revision: D62671994