remove torch.cuda.synchronize() from run_nvfp4_grouped_gemm (cudagraph-safe)

2026-05-16 20:42:49 +00:00
parent 5e245bc0c6
commit 3465b9d471
1 changed files with 0 additions and 1 deletions
--- a/cutedsl/bridge.py
+++ b/cutedsl/bridge.py
@@ -344,6 +344,5 @@ def run_nvfp4_grouped_gemm(
        stream,
        global_scale_a=gsa_c, global_scale_b=gsb_c,
    )
-    torch.cuda.synchronize()

    return out