[Improve] Use PyTorch official scaled_dot_product_attention
to accelerate MultiheadAttention
.
#1434
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Outdated suggestions cannot be applied.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.
Motivation
Pytorch 2.0 has official
scaled_dot_product_attention
implementation, which will automatically select theattention implementation like
FlashAttention
andMemory-Efficient Attention
.Modification
If the PyTorch version is higher than 2.0, use the official
scaled_dot_product_attention
in theMultiheadAttention
.Use cases
Here is the speed comparsion on test (NVIDIA A100, FP16):
Original:
New (Use flash-attention):
Checklist
Before PR:
After PR: