disable LT interface #74021
Closed
disable LT interface #74021
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
CI Flow Status⚛️ CI FlowRuleset - Version:
|
Contributor
🔗 Helpful links
💊 CI failures summary and remediationsAs of commit 2ed8c19 (more details on the Dr. CI page): 💚 💚 Looks good so far! There are no failures yet. 💚 💚 This comment was automatically generated by Dr. CI (expand for details).Please report bugs/suggestions to the (internal) Dr. CI Users group. |
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D34775050 |
Summary: Pull Request resolved: pytorch#74021 Disables cublasLt as it is buggy with cuda 11.0. We'll reenable it based on cuda version (some known bugs are fixed in 11.5) Test Plan: Existing tests Reviewed By: jspark1105 Differential Revision: D34775050 fbshipit-source-id: 6931db465c98c452ec3cc49947de9daeb4ce0465
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D34775050 |
ngimel
force-pushed
the
export-D34775050
branch
from
514fa4a to
2ed8c19
Compare
Collaborator
Author
|
I'm guarding use of cublasLt on cuda version 11.4, and I've removed "batch size" condition, because we are doing matrix multiply (not batched matrix multiply), so batch limitation doesn't apply (this still might have some perf issues though, we also need to specify workspace) |
Collaborator
Collaborator
Author
|
Oh right, I missed it. @xwang233 if you know of any additional conditions that should be used to determine whether to use cublasLt, or if you could do more thorough testing with cuda 11.5 to make sure there's no regression in functionality and perf, that would be great. |
facebook-github-bot
pushed a commit
that referenced
this pull request
Mar 10, 2022
Summary: Pull Request resolved: #74021 Disables cublasLt as it is buggy with cuda 11.0. We'll reenable it based on cuda version (some known bugs are fixed in 11.5) Test Plan: Existing tests Reviewed By: jspark1105 Differential Revision: D34775050 fbshipit-source-id: 3faf3d15a0d9e24ca99e3ab515d2bc73a4c51fb5
Contributor
|
Hey @ngimel. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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: Disables cublasLt as it is buggy with cuda 11.0. We'll reenable it based on cuda version (some known bugs are fixed in 11.5)
Test Plan: Existing tests
Differential Revision: D34775050