ci: tag-gated nightly migration — foundation + 40 whole-file moves (#24725)

Co-authored-by: hnyls2002 <lsyincs@gmail.com>
Co-authored-by: Liangsheng Yin <hnyls2002@gmail.com>

test/registered/moe/test_cutedsl_moe.py

@@ -16,7 +16,7 @@ except ImportError:
     CuteDslMoEWrapper = None
     convert_sf_to_mma_layout = None
 
-register_cuda_ci(est_time=24, suite="nightly-4-gpu-b200", nightly=True)
+register_cuda_ci(est_time=24, stage="extra-b", runner_config="4-gpu-b200")
 
 SKIP_TEST = torch.cuda.get_device_capability() < (10, 0)
 SKIP_REASON = "Nvfp4 Requires compute capability of 10 or above."

test/registered/moe/test_moe_ep.py

@@ -12,6 +12,8 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+# Per-commit: TP=2 EP=2 baseline.
+# DeepGEMM/FP8 variant moved to test_moe_ep_nightly.py.
 register_cuda_ci(est_time=279, stage="stage-b", runner_config="2-gpu-large")
 
 
@@ -52,46 +54,5 @@ class TestEp(CustomTestCase):
         self.assertGreater(metrics["score"], 0.60)
 
 
-class TestEpDeepGEMM(CustomTestCase):
-    @classmethod
-    def setUpClass(cls):
-        cls.model = DEFAULT_MODEL_NAME_FOR_TEST_MLA
-        cls.base_url = DEFAULT_URL_FOR_TEST
-        cls.process = popen_launch_server(
-            cls.model,
-            cls.base_url,
-            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
-            other_args=[
-                "--trust-remote-code",
-                "--tp",
-                "2",
-                "--ep-size",
-                "2",
-                "--quantization",
-                "fp8",
-                "--moe-runner-backend",
-                "deep_gemm",
-            ],
-        )
-
-    @classmethod
-    def tearDownClass(cls):
-        kill_process_tree(cls.process.pid)
-
-    def test_gsm8k(self):
-        args = SimpleNamespace(
-            base_url=self.base_url,
-            eval_name="gsm8k",
-            api="completion",
-            max_tokens=512,
-            num_examples=200,
-            num_threads=128,
-        )
-        metrics = run_eval(args)
-        print(metrics)
-
-        self.assertGreater(metrics["score"], 0.60)
-
-
 if __name__ == "__main__":
     unittest.main()

test/registered/moe/test_moe_ep_extra.py

@@ -0,0 +1,66 @@
+"""Extra: TP=2 EP=2 with FP8 + DeepGEMM MoE backend.
+
+Sibling per-commit file (test_moe_ep.py) keeps the baseline TP=2 EP=2
+variant.
+"""
+
+import unittest
+from types import SimpleNamespace
+
+from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_cuda_ci
+from sglang.test.run_eval import run_eval
+from sglang.test.test_utils import (
+    DEFAULT_MODEL_NAME_FOR_TEST_MLA,
+    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+    DEFAULT_URL_FOR_TEST,
+    CustomTestCase,
+    popen_launch_server,
+)
+
+register_cuda_ci(est_time=279, stage="extra-a", runner_config="2-gpu-large")
+
+
+class TestEpDeepGEMM(CustomTestCase):
+    @classmethod
+    def setUpClass(cls):
+        cls.model = DEFAULT_MODEL_NAME_FOR_TEST_MLA
+        cls.base_url = DEFAULT_URL_FOR_TEST
+        cls.process = popen_launch_server(
+            cls.model,
+            cls.base_url,
+            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+            other_args=[
+                "--trust-remote-code",
+                "--tp",
+                "2",
+                "--ep-size",
+                "2",
+                "--quantization",
+                "fp8",
+                "--moe-runner-backend",
+                "deep_gemm",
+            ],
+        )
+
+    @classmethod
+    def tearDownClass(cls):
+        kill_process_tree(cls.process.pid)
+
+    def test_gsm8k(self):
+        args = SimpleNamespace(
+            base_url=self.base_url,
+            eval_name="gsm8k",
+            api="completion",
+            max_tokens=512,
+            num_examples=200,
+            num_threads=128,
+        )
+        metrics = run_eval(args)
+        print(metrics)
+
+        self.assertGreater(metrics["score"], 0.60)
+
+
+if __name__ == "__main__":
+    unittest.main()