Skip to content

TransformerEngine FP8 is slower & more memory intensive than FlashAttention FP16? #3713

TransformerEngine FP8 is slower & more memory intensive than FlashAttention FP16?

TransformerEngine FP8 is slower & more memory intensive than FlashAttention FP16? #3713

Triggered via issue August 23, 2024 18:45
@ptrendxptrendx
commented on #1119 2215fa5
Status Skipped
Total duration 3s
Artifacts

blossom-ci.yml

on: issue_comment
Authorization
0s
Authorization
Upload log
0s
Upload log
Vulnerability scan
0s
Vulnerability scan
Start ci job
0s
Start ci job
Fit to window
Zoom out
Zoom in