[TRTLLM-4880, TRTLLM-4595] Add soft logit capping in custom kernel and flashinfer #62
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
Signed-off-by: Chenghao Zhang <[email protected]>
Signed-off-by: Chenghao Zhang <[email protected]>
Signed-off-by: Chenghao Zhang <[email protected]>
lucaslie
approved these changes
Jun 26, 2025
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Outdated suggestions cannot be applied.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.
Description
As title, this PR adds the support for soft logitcapping introduced in the Gemma 2 paper. The paper - https://arxiv.org/pdf/2408.00118, the soft logitcapping is the calculation that applies before the softmax op. And the math is
: LOGIT_CAP * tanh(attn / LOGIT_CAP)
Test Coverage
Added two new tests to test the soft logit capping. test_gqa_op_with_logit_cap and test_flashinfer_attention_op_context_with_logit_cap