[ROCm] Update meta_registration for efficient attention (#146979) #2050
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
Fixes a series of failing and skipped unit tests. For nvidia hw, the longsumexp last dimension is required to be a multiple of 32. This is not the case for rocm. A related issue: pytorch#146848 The unit tests in question: ```bash inductor.test_fused_attention SDPAPatternRewriterCudaDynamicTests test_sdpa_prev_13_cuda inductor.test_fused_attention SDPAPatternRewriterCudaDynamicTests test_sdpa_prev_14_cuda inductor.test_fused_attention SDPAPatternRewriterCudaDynamicTests test_sdpa_prev_15_cuda inductor.test_fused_attention SDPAPatternRewriterCudaDynamicTests test_sdpa_rewriter_11_cuda inductor.test_fused_attention SDPAPatternRewriterCudaDynamicTests test_sdpa_rewriter_14_cuda inductor.test_fused_attention SDPAPatternRewriterCudaDynamicTests test_sdpa_rewriter_15_cuda inductor.test_fused_attention SDPAPatternRewriterCudaDynamicTests test_sdpa_rewriter_17_cuda inductor.test_fused_attention SDPAPatternRewriterCudaDynamicTests test_sdpa_rewriter_1_cuda inductor.test_fused_attention SDPAPatternRewriterCudaDynamicTests test_sdpa_rewriter_1_freezing inductor.test_fused_attention SDPAPatternRewriterCudaDynamicTests test_sdpa_rewriter_2_cuda inductor.test_fused_attention SDPAPatternRewriterCudaDynamicTests test_sdpa_rewriter_3_cuda inductor.test_fused_attention SDPAPatternRewriterCudaDynamicTests test_sdpa_rewriter_4_cuda inductor.test_fused_attention SDPAPatternRewriterCudaDynamicTests test_sdpa_rewriter_6_cuda inductor.test_fused_attention SDPAPatternRewriterCudaTests test_sdpa_prev_13_cuda inductor.test_fused_attention SDPAPatternRewriterCudaTests test_sdpa_prev_14_cuda inductor.test_fused_attention SDPAPatternRewriterCudaTests test_sdpa_prev_15_cuda inductor.test_fused_attention SDPAPatternRewriterCudaTests test_sdpa_rewriter_11_cuda inductor.test_fused_attention SDPAPatternRewriterCudaTests test_sdpa_rewriter_14_cuda inductor.test_fused_attention SDPAPatternRewriterCudaTests test_sdpa_rewriter_15_cuda inductor.test_fused_attention SDPAPatternRewriterCudaTests test_sdpa_rewriter_17_cuda inductor.test_fused_attention SDPAPatternRewriterCudaTests test_sdpa_rewriter_1_cuda inductor.test_fused_attention SDPAPatternRewriterCudaTests test_sdpa_rewriter_1_freezing inductor.test_fused_attention SDPAPatternRewriterCudaTests test_sdpa_rewriter_2_cuda inductor.test_fused_attention SDPAPatternRewriterCudaTests test_sdpa_rewriter_3_cuda inductor.test_fused_attention SDPAPatternRewriterCudaTests test_sdpa_rewriter_4_cuda inductor.test_fused_attention SDPAPatternRewriterCudaTests test_sdpa_rewriter_6_cuda ``` Pull Request resolved: pytorch#146979 Approved by: https://github.com/shunting314
|
Jenkins build for e13a75c5da6d5a57a9c0d66f5b47e5cede0f8ed3 commit is in progress |
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.
Fixes a series of failing and skipped unit tests.
For nvidia hw, the longsumexp last dimension is required to be a multiple of 32. This is not the case for rocm.
A related issue: pytorch#146848
The unit tests in question:
Pull Request resolved: pytorch#146979
Approved by: https://github.com/shunting314
Fixes #SWDEV-528386