Commit ee941ac

authored

[https://nvbugs/5456493][feat] add fp8 dense for sm120 (#9174)

Signed-off-by: CarstyYou <[email protected]>

1 parent a79c0df commit ee941acCopy full SHA for ee941ac

6 files changed

+969

-8

lines changed

cpp/tensorrt_llm
- kernels/cutlass_kernels/fp8_blockscale_gemm
  - 6kd_blockwise_gemm
    - sm120_fp8_gemm_1d2d.cuh
    - sm120_utils.cuh
  - fp8_blockscale_gemm_kernel.cuh
- thop
  - fp8BlockScalingGemm.cpp
tensorrt_llm/_torch/modules
- linear.py
tests/unittest/_torch/thop/parallel
- test_fp8_block_scale_gemm.py

6 files changed

+969

-8

lines changed

Comments

(0)