Enables Split-KV avoidance and updates error messages #74
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
Uncomments code that forces num_splits to 1 and resets accumulator tensors to avoid extra memory overhead from Split-KV operations. Updates error messages to reference "FlashDynamicMaskAttention" instead of "FlashAttention" for consistency with the library's actual implementation.
Contributor
There was a problem hiding this comment.
Pull Request Overview
This PR enables Split-KV avoidance by uncommenting code that forces the number of splits to 1 and resets accumulator tensors, while also updating error messages for consistency with the FlashDynamicMaskAttention library implementation.
- Uncomments code to force
num_splits = 1and reset accumulator tensors to avoid Split-KV memory overhead - Updates error messages to reference "FlashDynamicMaskAttention" instead of "FlashAttention"
| auto q_dtype = q.dtype(); | ||
| TORCH_CHECK(q_dtype == torch::kFloat16 || q_dtype == torch::kBFloat16, | ||
| "FlashAttention only support fp16 and bf16 data type"); | ||
| TORCH_CHECK(q_dtype == torch::kFloat16 || q_dtype == torch::kBFloat16, "FlashDynamicMaskAttention only support fp16 and bf16 data type"); |
There was a problem hiding this comment.
Grammar error: "only support" should be "only supports" to maintain subject-verb agreement.
Suggested change
| TORCH_CHECK(q_dtype == torch::kFloat16 || q_dtype == torch::kBFloat16, "FlashDynamicMaskAttention only support fp16 and bf16 data type"); | |
| TORCH_CHECK(q_dtype == torch::kFloat16 || q_dtype == torch::kBFloat16, "FlashDynamicMaskAttention only supports fp16 and bf16 data type"); |
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.
Uncomments code that forces num_splits to 1 and resets accumulator tensors to avoid extra memory overhead from Split-KV operations.
Updates error messages to reference "FlashDynamicMaskAttention" instead of "FlashAttention" for consistency with the library's actual implementation.