Fix persistent compilation cache not working on TPU#9759
Open
shr1ram wants to merge 1 commit into
Open
Conversation
The TPU backend (via PjRtCApiClient) implements LoadSerializedExecutable but does not override DeserializeExecutable, which returns UNIMPLEMENTED from the base PjRtClient class. This caused every cache load attempt to fail, forcing recompilation on every run even when valid cached executables existed on disk. Fix by trying LoadSerializedExecutable first (the path implemented by PJRT C API plugins like TPU), then falling back to the two-step DeserializeExecutable + Load path for backends that implement that instead. Fixes pytorch#9094 Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
qihqi
approved these changes
Apr 2, 2026
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
DeserializeComputationcallsclient_->DeserializeExecutable(), which returnsUNIMPLEMENTEDon TPU because the TPU backend (PjRtCApiClient) implementsLoadSerializedExecutableinsteadLoadSerializedExecutablefirst (the PJRT C API plugin path used by TPU/Neuron), then falls back to the oldDeserializeExecutable+Loadtwo-step path for other backendsFixes #9094
Test plan
test_persistent_cache.pytests should continue to pass on TPUPersistentCacheHitmetric) and significantly faster startup🤖 Generated with Claude Code