fix(ModelStage): del self.model before gc.collect() in teardown()#1965
Draft
rlratzel wants to merge 3 commits into
Draft
fix(ModelStage): del self.model before gc.collect() in teardown()#1965rlratzel wants to merge 3 commits into
rlratzel wants to merge 3 commits into
Conversation
Explicitly delete the model reference before calling gc.collect() so the garbage collector can actually free the model weights. Without this, the reference remains live and gc.collect() has nothing to reclaim, leaving GPU memory occupied after the benchmark subprocess exits. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> Signed-off-by: rlratzel <rratzel@nvidia.com>
|
Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually. Contributors can view more details about this message here. |
Signed-off-by: rlratzel <rratzel@nvidia.com>
Signed-off-by: rlratzel <rratzel@nvidia.com>
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Fixes GPU memory not being released after xenna-based benchmarks complete, causing subsequent benchmarks to fail with out-of-memory errors.
ModelStage.teardown()was callinggc.collect()without first dropping the reference toself.model. Since the model was still referenced, the garbage collector had nothing to reclaim, leaving the model weights in GPU memory after the benchmark subprocess exited.The fix explicitly deletes
self.modelbeforegc.collect()so the reference count drops to zero and the memory is actually freed beforetorch.cuda.empty_cache()runs. Ahasattrguard is included for safety in case teardown is called before setup completes.Status
Draft — needs to be verified against benchmark logs and a test run on the affected machine before merging.
Test plan
🤖 Generated with Claude Code