Rename sglang.bench_latency to sglang.bench_one_batch (#2118)

test/srt/test_bench_one_batch.py

@@ -4,19 +4,19 @@ from sglang.test.test_utils import (
     DEFAULT_MODEL_NAME_FOR_TEST,
     DEFAULT_MOE_MODEL_NAME_FOR_TEST,
     is_in_ci,
-    run_bench_latency,
+    run_bench_one_batch,
 )
 
 
-class TestBenchLatency(unittest.TestCase):
+class TestBenchOneBatch(unittest.TestCase):
     def test_default(self):
-        output_throughput = run_bench_latency(DEFAULT_MODEL_NAME_FOR_TEST, [])
+        output_throughput = run_bench_one_batch(DEFAULT_MODEL_NAME_FOR_TEST, [])
 
         if is_in_ci():
             self.assertGreater(output_throughput, 135)
 
     def test_moe_default(self):
-        output_throughput = run_bench_latency(
+        output_throughput = run_bench_one_batch(
             DEFAULT_MOE_MODEL_NAME_FOR_TEST, ["--tp", "2"]
         )
 

test/srt/test_torch_tp.py

@@ -1,11 +1,11 @@
 import unittest
 
-from sglang.test.test_utils import is_in_ci, run_bench_latency
+from sglang.test.test_utils import is_in_ci, run_bench_one_batch
 
 
 class TestTorchTP(unittest.TestCase):
     def test_torch_native_llama(self):
-        output_throughput = run_bench_latency(
+        output_throughput = run_bench_one_batch(
             "meta-llama/Meta-Llama-3-8B",
             [
                 "--tp",

test/srt/test_triton_attention_backend.py

@@ -14,13 +14,13 @@ from sglang.test.test_utils import (
     DEFAULT_URL_FOR_TEST,
     is_in_ci,
     popen_launch_server,
-    run_bench_latency,
+    run_bench_one_batch,
 )
 
 
 class TestTritonAttnBackend(unittest.TestCase):
     def test_latency(self):
-        output_throughput = run_bench_latency(
+        output_throughput = run_bench_one_batch(
             DEFAULT_MODEL_NAME_FOR_TEST,
             [
                 "--attention-backend",