model: Support Hybrid Mamba2 NemotronHForCausalLM (nvidia/NVIDIA-Nemotron-Nano-9B-v2) (#10909)

Signed-off-by: Netanel Haber <nhaber@nvidia.com>

test/srt/models/test_generation_models.py

@@ -91,6 +91,11 @@ ALL_MODELS = [
         trust_remote_code=True,
         skip_long_prompt=True,
     ),
+    ModelCase(
+        "nvidia/NVIDIA-Nemotron-Nano-9B-v2",
+        trust_remote_code=True,
+        skip_long_prompt=True,
+    ),
     ModelCase(
         "swiss-ai/Apertus-8B",
         trust_remote_code=True,

test/srt/models/test_nvidia_nemotron_nano_v2.py

@@ -0,0 +1,44 @@
+from types import SimpleNamespace
+
+from sglang.srt.utils import kill_process_tree
+from sglang.test.few_shot_gsm8k import run_eval
+from sglang.test.test_utils import (
+    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+    DEFAULT_URL_FOR_TEST,
+    CustomTestCase,
+    popen_launch_server,
+)
+
+
+class TestNvidiaNemotronNanoV2(CustomTestCase):
+    @classmethod
+    def setUpClass(cls):
+        cls.model = "nvidia/NVIDIA-Nemotron-Nano-9B-v2"
+        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=[
+                "--max-mamba-cache-size",
+                "256",
+            ],
+        )
+
+    @classmethod
+    def tearDownClass(cls):
+        kill_process_tree(cls.process.pid)
+
+    def test_gsm8k(self):
+        args = SimpleNamespace(
+            num_shots=5,
+            data_path=None,
+            num_questions=200,
+            max_new_tokens=512,
+            parallel=128,
+            host="http://127.0.0.1",
+            port=int(self.base_url.split(":")[-1]),
+        )
+        metrics = run_eval(args)
+        print(f"{metrics=}")
+        self.assertGreater(metrics["accuracy"], 0.87)