Fix FlashAttention3 optimal use when available#195
Open
3manifold wants to merge 1 commit intoLightricks:mainfrom
Open
Fix FlashAttention3 optimal use when available#1953manifold wants to merge 1 commit intoLightricks:mainfrom
3manifold wants to merge 1 commit intoLightricks:mainfrom
Conversation
3manifold
changed the title
3manifold
changed the title
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
When the model cfg is set to
"attention_type": "default"(e.g. seeconfig {"transformer": {"_class_name": "AVTransformer3DModel",...,"attention_type": "default", ...} }in https://huggingface.co/Lightricks/LTX-2.3/blob/main/ltx-2.3-22b-dev.safetensors),AttentionCallableis not treated optimally inpackages/ltx-core/src/ltx_core/model/transformer/attention.py.In brief,
packages/ltx-core/src/ltx_core/model/transformer/model_configurator.pypasses downattention_typetopackages/ltx-core/src/ltx_core/model/transformer/attention.py. There, even if the imports flag the various attention cases correctly (seememory_efficient_attention,flash_attn_interface) at the top of the file,class AttentionFunction(Enum)is mishandlingAttentionCallablevalues.This PR fixes that behaviour following the principle of
FA3 > XFormers > PyTorch. It also adds a fallback fromFlashAttention3toPytorchAttentionin case of arbitrary attention masks to avoid errors during runs.resolves #196