Check size on transformer cache #942
Open
Conversation
This file contains 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
4 tasks
graemenail
force-pushed
the
fix-transformer-cache
branch
from
a817edd
to
d186d94
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Description
This PR reinstates a shape check on the entries of transformer cache. This PR is the minimal change required to address the issue highlighted in #881.
#881 was reported to have increased memory usage in normal operation. In that PR, the tensor shape was added to the cache key for better retrieval. However, this change led to many more cache entries and because this cache is a member of the Transformer, the lifetime of cache entries were tied to that of the Transformer and ultimately persisted too long.
Related: #881
List of changes:
Added dependencies: none
How to test
Checklist