Commits on Oct 9, 2024

1. PR #15331: Support cuDNN frontend scaled dot product attention for FP… …

   …8. Part- 2(backward)
   
   Imported from GitHub PR #15331
   
   As the 2nd part of #15092.
   NOTE: this feature relies on cudnn-frontend v1.6.1 which is not in XLA yet.
   Copybara import of the project:
   
   --
   06db3c8 by shuw <shuw@nvidia.com>:
   
   Scaled dot product attention implementation by cudnn.
   
   --
   937b0e2 by shuw <shuw@nvidia.com>:
   
   Improve after review 1
   
   --
   398b2ba by shuw <shuw@nvidia.com>:
   
   clang-format
   
   --
   0825789 by Shu Wang <shuw@nvidia.com>:
   
   fix typo.
   --
   d0ae3cf by shuw <shuw@nvidia.com>:
   
   Refactor test
   
   Merging this change closes #15331
   
   COPYBARA_INTEGRATE_REVIEW=#15331 from wenscarl:sdpa_fp8_bwd d0ae3cf
   PiperOrigin-RevId: 684062495

   

   wenscarl authored and Google-ML-Automation committed Oct 9, 2024
   Configuration menu

   • View commit details
   • Copy full SHA for 467563e
   • Browse repository at this point

   Copy the full SHA

   467563e View commit details

   Browse the repository at this point in the history