[2.0.1] Disable SDPA FlashAttention backward and mem eff attention on sm86+ for head_dim above 64 #99736
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
cpuhrsch
force-pushed
the
sdpasm86fix201
branch
from
cf3534a
to
cb45ca9
Compare
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/99736
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ⏳ No Failures, 2 PendingAs of commit d3c53c4: This comment was automatically generated by Dr. CI and updates every 15 minutes. |
cpuhrsch
force-pushed
the
sdpasm86fix201
branch
from
cb45ca9
to
6c31735
Compare
cpuhrsch
force-pushed
the
sdpasm86fix201
branch
from
6c31735
to
85810e3
Compare
malfet
approved these changes
Apr 21, 2023
.github/workflows/pull.yml
Outdated
Comment on lines
369
to
392
|
|
||
| linux-bionic-cuda11_8-py3_10-gcc7-sm89-build: | ||
| name: linux-bionic-cuda11.8-py3.10-gcc7-sm89 | ||
| uses: ./.github/workflows/_linux-build.yml | ||
| with: | ||
| build-environment: linux-bionic-cuda11.8-py3.10-gcc7-sm89 | ||
| docker-image-name: pytorch-linux-bionic-cuda11.8-cudnn8-py3-gcc7 | ||
| cuda-arch-list: '8.9' | ||
| test-matrix: | | ||
| { include: [ | ||
| { config: "default", shard: 1, num_shards: 4, runner: "linux.gcp.l4" }, | ||
| { config: "default", shard: 2, num_shards: 4, runner: "linux.gcp.l4" }, | ||
| { config: "default", shard: 3, num_shards: 4, runner: "linux.gcp.l4" }, | ||
| { config: "default", shard: 4, num_shards: 4, runner: "linux.gcp.l4" }, | ||
| ]} | ||
| linux-bionic-cuda11_8-py3_10-gcc7-sm89-test: | ||
| name: linux-bionic-cuda11.8-py3.10-gcc7-sm89 | ||
| uses: ./.github/workflows/_linux-test.yml | ||
| needs: linux-bionic-cuda11_8-py3_10-gcc7-sm89-build | ||
| with: | ||
| build-environment: linux-bionic-cuda11.8-py3.10-gcc7-sm89 | ||
| docker-image: ${{ needs.linux-bionic-cuda11_8-py3_10-gcc7-sm89-build.outputs.docker-image }} | ||
| test-matrix: ${{ needs.linux-bionic-cuda11_8-py3_10-gcc7-sm89-build.outputs.test-matrix }} | ||
| use-gha: anything-non-empty-to-use-gha |
There was a problem hiding this comment.
This needs to be removed before cherry-pick is landed
There was a problem hiding this comment.
Oh I thought you wanted to include this so we can get sm89 tests on release branch. I'll remove it.
…or head_dim above 64 (pytorch#99105) Expand sdpa_utils.h check to disable FlashAttention when using autograd and mem eff attention for the following cases - head_dim > 64 - sm86 or newer Previously we only disable these kernels on sm86 and for head_dim equal to 128. Pull Request resolved: pytorch#99105 Approved by: https://github.com/malfet
|
The new workflow has been removed, rebased and force pushed. |
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.
2.0.1 submission of #99105 plus sm89 CI enablement.
Expand sdpa_utils.h check to disable FlashAttention when using autograd and mem eff attention for the following cases
Previously we only disable these kernels on sm86 and for head_dim equal to 128.