[CI] Add CI test for DeepSeek V3.2 MTP (#11835)

python/sglang/srt/server_args.py

@@ -1201,6 +1201,9 @@ class ServerArgs:
                 )
             if self.max_running_requests is None:
                 self.max_running_requests = 48
+                logger.warning(
+                    "Max running requests is reset to 48 for speculative decoding."
+                )
 
             if self.speculative_algorithm == "EAGLE" and self.enable_beta_spec:
                 self.disable_overlap_schedule = False

test/srt/run_suite.py

@@ -181,6 +181,7 @@ suites = {
     ],
     "per-commit-8-gpu-h200-deepseek-v32": [
         TestFile("test_deepseek_v32_basic.py", 275),
+        TestFile("test_deepseek_v32_mtp.py", 275),
     ],
     "vllm_dependency_test": [
         TestFile("quant/test_awq.py", 163),

test/srt/test_deepseek_v32_basic.py

@@ -16,7 +16,7 @@ from sglang.test.test_utils import (
 DEEPSEEK_V32_MODEL_PATH = "deepseek-ai/DeepSeek-V3.2-Exp"
 
 
-class TestDeepseekV3Basic(CustomTestCase):
+class TestDeepseekV32Basic(CustomTestCase):
     @classmethod
     def setUpClass(cls):
         cls.model = DEEPSEEK_V32_MODEL_PATH
@@ -57,7 +57,7 @@ class TestDeepseekV3Basic(CustomTestCase):
 
         if is_in_ci():
             write_github_step_summary(
-                f"### test_gsm8k (deepseek-v3)\n" f'{metrics["accuracy"]=:.3f}\n'
+                f"### test_gsm8k (deepseek-v32)\n" f'{metrics["accuracy"]=:.3f}\n'
             )
             self.assertGreater(metrics["accuracy"], 0.935)
 
@@ -69,7 +69,7 @@ class TestDeepseekV3Basic(CustomTestCase):
 
         if is_in_ci():
             write_github_step_summary(
-                f"### test_bs_1_speed (deepseek-v3)\n" f"{speed=:.2f} token/s\n"
+                f"### test_bs_1_speed (deepseek-v32)\n" f"{speed=:.2f} token/s\n"
             )
             self.assertGreater(speed, 50)
 

test/srt/test_deepseek_v32_mtp.py

+import unittest
+from types import SimpleNamespace
+
+import requests
+
+from sglang.srt.utils import kill_process_tree
+from sglang.test.few_shot_gsm8k import run_eval as run_eval_few_shot_gsm8k
+from sglang.test.send_one import BenchArgs, send_one_prompt
+from sglang.test.test_utils import (
+    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+    DEFAULT_URL_FOR_TEST,
+    CustomTestCase,
+    is_in_ci,
+    popen_launch_server,
+    write_github_step_summary,
+)
+
+FULL_DEEPSEEK_V3_MODEL_PATH = "deepseek-ai/DeepSeek-V3.2-Exp"
+
+
+class TestDeepseekV32MTP(CustomTestCase):
+    @classmethod
+    def setUpClass(cls):
+        cls.model = FULL_DEEPSEEK_V3_MODEL_PATH
+        cls.base_url = DEFAULT_URL_FOR_TEST
+        other_args = [
+            "--trust-remote-code",
+            "--tp",
+            "8",
+            "--dp",
+            "8",
+            "--enable-dp-attention",
+            "--speculative-algorithm",
+            "EAGLE",
+            "--speculative-num-steps",
+            "3",
+            "--speculative-eagle-topk",
+            "1",
+            "--speculative-num-draft-tokens",
+            "4",
+            "--mem-frac",
+            "0.7",
+        ]
+        cls.process = popen_launch_server(
+            cls.model,
+            cls.base_url,
+            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+            other_args=other_args,
+        )
+
+    @classmethod
+    def tearDownClass(cls):
+        kill_process_tree(cls.process.pid)
+
+    def test_a_gsm8k(
+        self,
+    ):  # Append an "a" to make this test run first (alphabetically) to warm up the server
+        requests.get(self.base_url + "/flush_cache")
+
+        args = SimpleNamespace(
+            num_shots=8,
+            data_path=None,
+            num_questions=1400,
+            parallel=1400,
+            max_new_tokens=512,
+            host="http://127.0.0.1",
+            port=int(self.base_url.split(":")[-1]),
+        )
+        metrics = run_eval_few_shot_gsm8k(args)
+        print(f"{metrics=}")
+
+        server_info = requests.get(self.base_url + "/get_server_info")
+        avg_spec_accept_length = server_info.json()["internal_states"][0][
+            "avg_spec_accept_length"
+        ]
+        print(f"{avg_spec_accept_length=}")
+
+        if is_in_ci():
+            write_github_step_summary(
+                f"### test_gsm8k (deepseek-v32 mtp)\n"
+                f'{metrics["accuracy"]=:.3f}\n'
+                f"{avg_spec_accept_length=:.2f}\n"
+            )
+            self.assertGreater(metrics["accuracy"], 0.935)
+            self.assertGreater(avg_spec_accept_length, 2.9)
+
+    def test_bs_1_speed(self):
+        args = BenchArgs(port=int(self.base_url.split(":")[-1]), max_new_tokens=2048)
+        acc_length, speed = send_one_prompt(args)
+
+        print(f"{acc_length=:.2f} {speed=:.2f}")
+
+        if is_in_ci():
+            write_github_step_summary(
+                f"### test_bs_1_speed (deepseek-v32 mtp)\n"
+                f"{acc_length=:.2f}\n"
+                f"{speed=:.2f} token/s\n"
+            )
+
+            self.assertGreater(acc_length, 2.9)
+            self.assertGreater(speed, 75)
+
+
+if __name__ == "__main__":
+    unittest.main()