onnxruntime
[JS/WebGPU] Avoid producing presentKey/presentValue outputs if pastKey/pastValue …
#21782
Merged

[JS/WebGPU] Avoid producing presentKey/presentValue outputs if pastKey/pastValue … #21782

satyajandhyala
satyajandhyala Avoid producing presentKey/presentValue outputs if pastKey/pastValue …
18a39066
satyajandhyala satyajandhyala marked this pull request as ready for review 1 year ago
satyajandhyala satyajandhyala requested a review from fs-eire fs-eire 1 year ago
tianleiwu
satyajandhyala
satyajandhyala
satyajandhyala satyajandhyala added ep:WebGPU
satyajandhyala Don't treat empty inputs as undefined in MHA. Let Attention deal with…
9e225640
satyajandhyala satyajandhyala marked this pull request as draft 1 year ago
satyajandhyala satyajandhyala requested a review from guschmue guschmue 1 year ago
satyajandhyala Feed pastKey/pastValue inputs down to the functions that generate sha…
c01f721c
satyajandhyala Added a test case with zero-size pastKey/pastValue inputs that requir…
c8b187ff
satyajandhyala Added back the assumption comment.
d0b0627b
satyajandhyala
satyajandhyala satyajandhyala marked this pull request as ready for review 1 year ago
satyajandhyala ShapeUtis.size should return 0 if the tensor dims is empty instead of 1.
d8aeee17
satyajandhyala Revert "ShapeUtis.size should return 0 if the tensor dims is empty in…
7520dc24
tianleiwu
tianleiwu commented on 2024-08-19
satyajandhyala Reverted changes preventing shape error checking.
c0d72a4e
satyajandhyala satyajandhyala requested a review from tianleiwu tianleiwu 1 year ago
tianleiwu
guschmue
guschmue
guschmue
guschmue approved these changes on 2024-08-19
satyajandhyala satyajandhyala merged 1fb2e71d into main 1 year ago
satyajandhyala satyajandhyala deleted the sajandhy/wepu_fix_attention_output branch 1 year ago

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone