update

test/srt/test_overlap_schedule.py

@@ -0,0 +1,52 @@
+import unittest
+from types import SimpleNamespace
+
+from sglang.srt.utils import kill_child_process
+from sglang.test.run_eval import run_eval
+from sglang.test.test_utils import (
+    DEFAULT_MODEL_NAME_FOR_TEST,
+    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+    DEFAULT_URL_FOR_TEST,
+    popen_launch_server,
+)
+
+
+class TestChunkedPrefill(unittest.TestCase):
+    def run_mmlu(self, disable_radix_cache, chunked_prefill_size=32):
+        other_args = ["--chunked-prefill-size", str(chunked_prefill_size)]
+        if disable_radix_cache:
+            other_args += ["--disable-radix-cache"]
+        other_args += ["--enable-overlap-schedule"]
+
+        model = DEFAULT_MODEL_NAME_FOR_TEST
+        base_url = DEFAULT_URL_FOR_TEST
+        process = popen_launch_server(
+            model,
+            base_url,
+            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+            other_args=other_args,
+        )
+
+        args = SimpleNamespace(
+            base_url=base_url,
+            model=model,
+            eval_name="mmlu",
+            num_examples=64,
+            num_threads=32,
+        )
+
+        try:
+            metrics = run_eval(args)
+            assert metrics["score"] >= 0.65
+        finally:
+            kill_child_process(process.pid)
+
+    def test_no_radix_attention_no_chunked_prefill(self):
+        self.run_mmlu(disable_radix_cache=True, chunked_prefill_size=-1)
+
+    def test_radix_attention_no_chunked_prefill(self):
+        self.run_mmlu(disable_radix_cache=True, chunked_prefill_size=-1)
+
+
+if __name__ == "__main__":
+    unittest.main()