onnxruntime
Cuda: Decoder Masked Multihead Attention Q values get corrupted when using cross attention
#16721

Merged

Cuda: Decoder Masked Multihead Attention Q values get corrupted when using cross attention #16721

RyanUnderhill merged 1 commit into main from ryanunderhill/mmha_q_fix

This was accidentally moved into the `if(!params.is_cross_attention)`…

33c94854

RyanUnderhill requested a review from

gh-yewang 3 years ago

gh-yewang approved these changes on 2023-07-14

RyanUnderhill merged 2cf31a20 into main 3 years ago

RyanUnderhill deleted the ryanunderhill/mmha_q_fix branch 3 years ago

Reviewers

gh-yewang

Assignees

No one assigned

Labels

None yet

Milestone

No milestone