This repository was archived by the owner on Nov 17, 2023. It is now read-only.
[v1.9.x] Enable oneDNN BRGEMM kernel for FullyConnected #20591
Closed
+27
−10
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
added 3 commits
September 17, 2021 07:49
…f FC (apache#20450) * Add flag for disabling oneDNN BRGEMM implementation of FC * Review fixes * Update env_var.md
* Enable brgemm based on input info * fix sanity * Review fixes * Change function name * Fix typo * Align variable assignments * Fix review * use const reference
|
Hey @bgawrych , Thanks for submitting the PR
CI supported jobs: [edge, centos-gpu, windows-gpu, clang, unix-cpu, miscellaneous, website, centos-cpu, sanity, windows-cpu, unix-gpu] Note: |
mseth10
added
pr-awaiting-testing
mozga-intel
approved these changes
Sep 20, 2021
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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Outdated suggestions cannot be applied.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 backport changes from v1.x branch (#20533, #20450) and enables MXNet to use BRGEMM implementation of oneDNN inner_product (FullyConnected).
For large FullyConnected layers, like in BERT-Large, this implementation can speedup whole model execution for about 40%.
Relative results from EC2 M6i instance:
This change only makes sense if oneDNN will be updated (#20590)