Skip to content

GitLab

Unverified Commit 1eb8eade authored by yizhang2077's avatar yizhang2077 Committed by GitHub
Browse files

add control for cutlass fp8 blockwise gemm (#3727)

parent 3c7bfd7e
Hide whitespace changes
Inline Side-by-side
Showing with 3 additions and 0 deletions
python/sglang/srt/layers/quantization/fp8_utils.py
View file @ 1eb8eade
import os
from typing import List, Optional, Tuple from typing import List, Optional, Tuple
import torch import torch
...@@ -40,6 +41,8 @@ def normalize_e4m3fn_to_e4m3fnuz( ...@@ -40,6 +41,8 @@ def normalize_e4m3fn_to_e4m3fnuz(
def cutlass_block_fp8_supported() -> bool: def cutlass_block_fp8_supported() -> bool:
if os.environ.get("SUPPORT_CUTLASS_BLOCK_FP8") is None:
return False
if _is_cuda: if _is_cuda:
major, minor = torch.cuda.get_device_capability() major, minor = torch.cuda.get_device_capability()
sm_version = major * 10 + minor sm_version = major * 10 + minor
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment