biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

dynamic distpatch of fp8 kernels (#14245)

Browse Source 
Signed-off-by: Jeff Daily <jeff.daily@amd.com>
This commit is contained in:
Jeff Daily
2025-03-11 07:54:56 -07:00
committed by GitHub
parent 08a1a1121d
commit a1c8f3796c
25 changed files with 292 additions and 159 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
benchmarks/kernels/benchmark_moe.py
View File

@@ -18,8 +18,7 @@ from vllm.model_executor.layers.fused_moe.fused_moe import *
from vllm.platforms import current_platform
from vllm.utils import FlexibleArgumentParser
FP8_DTYPE = torch.float8_e4m3fnuz if current_platform.is_rocm(
) else torch.float8_e4m3fn
FP8_DTYPE = current_platform.fp8_dtype()
class BenchmarkConfig(TypedDict):