Commits on May 15, 2024

1. Support FP8 scale calculation with scalar and cleanup …

   Summary:
   Follow up on D57263833 to support FP8 scale calculation with scalar and merge two FP8 tensorwise GEMMs into one
   
   Note that besides `Sm90ScalarBroadcast` in CUTLASS, AMD CK f8f8bf16 GEMM also requires passing scales as scalar instead of tensor scalar. This support is required in both NV and AMD sides
   
   Differential Revision: D57367680

   

   jiawenliu64 authored and facebook-github-bot committed May 15, 2024
   Configuration menu

   • View commit details
   • Copy full SHA for 0cf43d9
   • Browse repository at this point

   Copy the full SHA

   0cf43d9 View commit details

   Browse the repository at this point in the history