EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

fix typo in deep gemm benchmarking(#3991)

Browse Source
This commit is contained in:
Xiaoyu Zhang
2025-03-02 16:34:00 +08:00
committed by GitHub
parent 90a55e2566
commit 50f28f65a0
1 changed files with 0 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
benchmark/kernels/deepseek/benchmark_deepgemm_fp8_gemm.py
View File

@@ -211,7 +211,6 @@ def get_benchmark(tp_size):
x = torch.randn((m, k), device="cuda", dtype=torch.bfloat16) x = torch.randn((m, k), device="cuda", dtype=torch.bfloat16)
y = torch.randn((n, k), device="cuda", dtype=torch.bfloat16) y = torch.randn((n, k), device="cuda", dtype=torch.bfloat16)
# 预处理数据，在计时之前完成
x_fp8, x_scale = per_token_cast_to_fp8(x) x_fp8, x_scale = per_token_cast_to_fp8(x)
y_fp8, y_scale = per_block_cast_to_fp8(y) y_fp8, y_scale = per_block_cast_to_fp8(y)
x_scale_col_major = get_col_major_tma_aligned_tensor(x_scale.clone()) x_scale_col_major = get_col_major_tma_aligned_tensor(x_scale.clone())