Fix Evoformer arch filtering consistency for mixed targets (#7863)#7872
Open
tohtana wants to merge 1 commit intodeepspeedai:masterfrom
Open
Fix Evoformer arch filtering consistency for mixed targets (#7863)#7872tohtana wants to merge 1 commit intodeepspeedai:masterfrom
tohtana wants to merge 1 commit intodeepspeedai:masterfrom
Conversation
…ai#7863) Normalize DS_EVOFORMER_GPU_ARCH values and reuse that normalization for both -DGPU_ARCH emission and Evoformer compute-capability filtering. This enforces a single Evoformer kernel-family floor during build and prunes lower cross-family TORCH_CUDA_ARCH_LIST entries. Add unit coverage for normalization/filter behavior and update the Evoformer tutorial to clarify the distinct roles of TORCH_CUDA_ARCH_LIST and DS_EVOFORMER_GPU_ARCH. Signed-off-by: Masahiro Tanaka <mtanaka@anyscale.com>
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 #7863
The root cause of #7863 was a compile-target consistency gap:
DS_EVOFORMER_GPU_ARCHselected one Evoformer family, whileTORCH_CUDA_ARCH_LISTcould still emit lower-family slices.The fix adds shared arch normalization in
EvoformerAttnBuilderand uses it in bothnvcc_args()(-DGPU_ARCH) andfilter_ccs()(-gencodepruning).For example, with
DS_EVOFORMER_GPU_ARCH=80,7.xtargets are now filtered out for Evoformer builds.This PR also adds unit tests to validate the new behavior.