Replace time.time() to time.perf_counter() for benchmarking. (#6178)

Signed-off-by: Lifu Huang <lifu.hlf@gmail.com>

benchmark/llm_judge/bench_other.py

@@ -85,7 +85,7 @@ def main(args):
     call_generate = partial(get_call_generate(args), temperature=0)
 
     # Run requests
-    tic = time.time()
+    tic = time.perf_counter()
 
     if args.backend != "lmql":
 
@@ -120,7 +120,7 @@ def main(args):
                 asyncio.gather(*[get_one_answer_async(i) for i in bt])
             )
 
-    latency = time.time() - tic
+    latency = time.perf_counter() - tic
 
     # Compute accuracy
     print(f"Latency: {latency:.3f}")

benchmark/llm_judge/bench_sglang.py

@@ -59,7 +59,7 @@ def main(args):
     backend = select_sglang_backend(args)
 
     # Run requests
-    tic = time.time()
+    tic = time.perf_counter()
     states = multi_dimension_judge.run_batch(
         arguments,
         temperature=0,
@@ -67,7 +67,7 @@ def main(args):
         num_threads=args.parallel,
         progress_bar=True,
     )
-    latency = time.time() - tic
+    latency = time.perf_counter() - tic
 
     print(f"Latency: {latency:.3f}")