Fix: Add out= parameter to run_fused_swiglu_grouped_gemm signature

2026-06-03 21:45:15 +00:00
parent 188ecae47f
commit 676fad064f
1 changed files with 1 additions and 0 deletions
--- a/dsv4/ops/gemm_runner.py
+++ b/dsv4/ops/gemm_runner.py
@@ -416,6 +416,7 @@ def run_fused_swiglu_grouped_gemm(
    swiglu_limit=0.0,
    mma_tiler_mn=(128, 128),
    cluster_shape_mn=(1, 1),
+    out=None,       # pre-allocated output buffer for CUDA graph capture
 ):
    """Run the fused SwiGLU NVFP4 scaled grouped GEMM.