From e252192679053bccc34ac7f291b2ba73533550f9 Mon Sep 17 00:00:00 2001 From: Ke Bao Date: Sat, 2 Aug 2025 06:37:59 +0800 Subject: [PATCH] Fix deepgemm masked grouped gemm jit compile (#8679) --- .../srt/layers/quantization/deep_gemm_wrapper/compile_utils.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/python/sglang/srt/layers/quantization/deep_gemm_wrapper/compile_utils.py b/python/sglang/srt/layers/quantization/deep_gemm_wrapper/compile_utils.py index 8949e3334..c3043f389 100644 --- a/python/sglang/srt/layers/quantization/deep_gemm_wrapper/compile_utils.py +++ b/python/sglang/srt/layers/quantization/deep_gemm_wrapper/compile_utils.py @@ -148,7 +148,7 @@ def _compile_grouped_gemm_nt_f8f8bf16_masked_one( "NUM_MATH_THREADS_PER_GROUP": num_math_threads_per_group, "N": n, "K": k, - "NUM_GROUPS": 1, + "NUM_GROUPS": num_groups, "BLOCK_M": block_m, "BLOCK_N": block_n, "BLOCK_K": block_k,