[Model] Add support for JetBrains' Mellum v2 code generation model (#27375)

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Co-authored-by: Jiminator <69131491+Jiminator@users.noreply.github.com>
Co-authored-by: Alex Nails <alex.nails@radixark.ai>

test/registered/unit/configs/test_model_config.py

@@ -0,0 +1,39 @@
+"""Unit tests for hybrid attention model configuration."""
+
+import unittest
+from types import SimpleNamespace
+
+from sglang.srt.configs.model_config import get_hybrid_layer_ids
+from sglang.test.ci.ci_register import register_cpu_ci
+from sglang.test.test_utils import CustomTestCase
+
+register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+
+
+class TestHybridLayerIds(CustomTestCase):
+    def test_layer_type_architectures(self):
+        config = SimpleNamespace(
+            num_hidden_layers=4,
+            layer_types=[
+                "sliding_attention",
+                "full_attention",
+                "sliding_attention",
+                "full_attention",
+            ],
+        )
+
+        for architecture in (
+            "Gemma4ForCausalLM",
+            "Gemma4ForConditionalGeneration",
+            "LagunaForCausalLM",
+            "MellumForCausalLM",
+        ):
+            with self.subTest(architecture=architecture):
+                self.assertEqual(
+                    get_hybrid_layer_ids([architecture], config),
+                    ([0, 2], [1, 3]),
+                )
+
+
+if __name__ == "__main__":
+    unittest.main()

test/registered/unit/models/test_mellum.py

@@ -0,0 +1,33 @@
+"""Unit tests for the Mellum model implementation."""
+
+import unittest
+
+import torch
+
+from sglang.srt.models.mellum import MellumForCausalLM
+from sglang.test.ci.ci_register import register_cpu_ci
+from sglang.test.test_utils import CustomTestCase
+
+register_cpu_ci(est_time=5, suite="base-a-test-cpu")
+
+
+class TestMellumForCausalLM(CustomTestCase):
+    def test_prepare_positions(self):
+        model = MellumForCausalLM.__new__(MellumForCausalLM)
+        positions = torch.tensor([[0, 1]], dtype=torch.int64).t()
+
+        model.use_fused_qk_norm_rope = False
+        self.assertIs(
+            model._prepare_positions(positions, torch.device("cpu")), positions
+        )
+
+        model.use_fused_qk_norm_rope = True
+        fused_positions = model._prepare_positions(positions, torch.device("cpu"))
+
+        self.assertEqual(fused_positions.dtype, torch.int32)
+        self.assertTrue(fused_positions.is_contiguous())
+        self.assertEqual(fused_positions.shape, (2,))
+
+
+if __name__ == "__main__":
+    unittest.main()