[KVCache] Support passing in attn_score_scaling_factor into KV cache #16606
Conversation
This file contains 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
|
cc: @MasterJH5574 Will need flashinfer-ai/flashinfer#126 to be merged first: |
yzh119
pushed a commit
to flashinfer-ai/flashinfer
that referenced
this pull request
Feb 19, 2024
In GPT-2, attention calculation requires an additional feature scale_attn_by_inverse_layer_idx. It provides a scaling factor per attention layer when calculating the attention score, before applying the softmax function. This PR supports this additional parameter in tvm_wrapper. See: apache/tvm#16606
|
Given flashinfer-ai/flashinfer#126 has been merged, let's bump 3rdparty/flashinfer to the latest FlashInfer |
|
Also there is a format issue https://ci.tlcpack.ai/blue/organizations/jenkins/tvm-lint/detail/PR-16606/3/pipeline |
MasterJH5574
approved these changes
Feb 20, 2024
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Outdated suggestions cannot be applied.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.
In GPT-2, attention calculation requires an additional feature
scale_attn_by_inverse_layer_idx. It provides a scaling factor per attention layer when calculating the attention score, before applying the softmax function.This PR supports this additional parameter in KV cache.