Fix bench script making input data on L2 cache (#7739)

sgl-kernel/benchmark/bench_per_token_group_quant_8bit.py

@@ -205,9 +205,9 @@ def benchmark(batch_size, seq_len, group_size, dst_dtype, provider):
     quantiles = [0.5, 0.2, 0.8]
 
     if provider == "triton":
-        fn = lambda: triton_per_token_group_quant_8bit(x.clone(), group_size, dst_dtype)
+        fn = lambda: triton_per_token_group_quant_8bit(x, group_size, dst_dtype)
     elif provider == "sglang":
-        fn = lambda: sglang_per_token_group_quant_8bit(x.clone(), group_size, dst_dtype)
+        fn = lambda: sglang_per_token_group_quant_8bit(x, group_size, dst_dtype)
 
     ms, min_ms, max_ms = triton.testing.do_bench(fn, quantiles=quantiles)