NVIDIA/cutlass v4.5.1

CUTLASS 4.5.1

on GitHub

CuTe DSL

• Bug fixing and improvements
  • Fixed following issues:
    #3219
    #3218
    #3212
    #3210
    #3208
    #3201
    #3227
  • Fixed Jax int64 stride divisibility issue
  • Fixed issues for SM120 blockscaled MMAs
    • added missing MXFP8MMAOP and MXF8F6F4MMAOP for sm120.

CUTLASS C++

• Fix SM100 F8F6F4 SS MMA (1SM and 2SM) traits to use typed op templates.
• Add UE8M0 (uniform exponent distribution) initialization support in tensor fill utilities.
• Add cvt.rn.bf16x2.e4m3x2 conversion instruction support to numeric_conversion.h.
• Update example 93 with paged KV cache support for Blackwell low-latency GQA.