diff --git a/benchmark/kernels/deepseek/benchmark_deepgemm_fp8_gemm.py b/benchmark/kernels/deepseek/benchmark_deepgemm_fp8_gemm.py
index db210b3a5..bcfd133f1 100644
--- a/benchmark/kernels/deepseek/benchmark_deepgemm_fp8_gemm.py
+++ b/benchmark/kernels/deepseek/benchmark_deepgemm_fp8_gemm.py
@@ -211,7 +211,6 @@ def get_benchmark(tp_size):
         x = torch.randn((m, 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)
         y_fp8, y_scale = per_block_cast_to_fp8(y)
         x_scale_col_major = get_col_major_tma_aligned_tensor(x_scale.clone())