Replace torch.cuda.Event with torch.Event for better hardware compatibility (#26985)

Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
2025-11-19 03:34:36 +08:00
parent c3e2978620
commit 2a2d5d2780
15 changed files with 41 additions and 48 deletions
--- a/benchmarks/kernels/benchmark_per_token_group_quant.py
+++ b/benchmarks/kernels/benchmark_per_token_group_quant.py
@@ -30,8 +30,8 @@ def _time_cuda(
        fn()
    torch.cuda.synchronize()

-    start = torch.cuda.Event(enable_timing=True)
-    end = torch.cuda.Event(enable_timing=True)
+    start = torch.Event(enable_timing=True)
+    end = torch.Event(enable_timing=True)

    start.record()
    for _ in range(bench_iters):