[Model] Add native MAGI-2 Preview diffusion support (#5918)

Signed-off-by: Hongsheng Liu <liuhongsheng4@huawei.com> Signed-off-by: hsliuustc0106 <liuhongsheng4@huawei.com>
2026-09-05 15:32:30 +08:00
1 parent 13c617a140
commit d4096e01fe
@@ -0,0 +1,77 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
+
+"""Released-checkpoint MAGI-2 Preview serving smoke on four H100 GPUs.
+
+This is a one-step 272p load/dispatch test, not a quality or performance test.
+It exercises the recommended resident SP4 topology through ``/v1/videos``.
+
+From ``tests/``::
+
+    pytest -s -v e2e/online_serving/test_magi2.py -m "advanced_model and diffusion" --run-level=advanced_model
+"""
+
+import json
+import os
+
+import pytest
+
+from tests.helpers.mark import hardware_marks
+from tests.helpers.runtime import OmniServer, OmniServerParams, OnlineOmniClient
+
+os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
+
+MODEL = "sand-ai/MAGI-2-preview"
+PROMPT = "A red fox walks through fresh snow while wind moves the pine branches."
+
+FOUR_CARD_MARKS = hardware_marks(res={"cuda": "H100"}, num_cards=4)
+
+
+@pytest.mark.core_model
+@pytest.mark.advanced_model
+@pytest.mark.diffusion
+@pytest.mark.parametrize(
+    "omni_server",
+    [
+        pytest.param(
+            OmniServerParams(
+                model=MODEL,
+                server_args=[
+                    "--model-class-name",
+                    "Magi2Pipeline",
+                    "--num-gpus",
+                    "4",
+                    "--tensor-parallel-size",
+                    "1",
+                    "--ulysses-degree",
+                    "4",
+                ],
+                init_timeout=1800,
+            ),
+            id="resident_sp4_272p_one_step",
+            marks=FOUR_CARD_MARKS,
+        )
+    ],
+    indirect=True,
+)
+def test_magi2_preview_serving_smoke(
+    omni_server: OmniServer,
+    online_client: OnlineOmniClient,
+) -> None:
+    request_config = {
+        "model": omni_server.model,
+        "form_data": {
+            "prompt": PROMPT,
+            "height": 256,
+            "width": 448,
+            "num_frames": 125,
+            "fps": 12.5,
+            "num_inference_steps": 1,
+            "seed": 42,
+            "extra_params": json.dumps({"resolution": "272p"}),
+        },
+        "expected_audio": {"sample_rate": 44100, "channels": 2},
+        "fps_tolerance": 0.01,
+    }
+
+    online_client.send_video_diffusion_request(request_config)