[CI/Build] Move L1/E2E coverage to weekly and split scheduled L4/L5 pipelines (#6311)

Signed-off-by: wangyu <410167048@qq.com>
Signed-off-by: [Your Name] <your.email@example.com>
Co-authored-by: amy-why-3459 <wuhaiyan17@huawei.com>

tests/buildkite/test_upload_pipeline.py

@@ -1,4 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
 from __future__ import annotations
 
 import sys
@@ -61,16 +62,23 @@ def test_bootstrap_steps_loaded_from_file() -> None:
 
 
 def test_docs_only_allows_main_scheduled_nightly_weekly_only() -> None:
-    """skip_all: no PR labels; main + NIGHTLY=1 / WEEKLY=1 still gates scheduled CI."""
+    """skip_all: no PR labels; main + NIGHTLY=1 / WEEKLY=1 / NON_CRITICAL=1 still gates scheduled CI."""
     rendered = _render(["docs/foo.md"])
     assert "key: image-build" in rendered
     assert "key: upload-nightly-pipeline" in rendered
     assert "key: upload-weekly-pipeline" in rendered
-    # Scheduled nightly also uploads L2/L3 with --e2e
-    assert "key: upload-ready-pipeline" in rendered
-    assert "key: upload-merge-pipeline" in rendered
-    assert 'build.env("NIGHTLY") == "1"' in rendered
+    # Scheduled main+WEEKLY=1 uploads L2/L3 with --e2e; NIGHTLY still gates L4 only.
+    doc = yaml.safe_load(rendered)
+    by_key = {step["key"]: step for step in doc["steps"]}
+    assert "NIGHTLY" not in by_key["upload-ready-pipeline"]["if"]
+    assert 'build.branch == "main"' in by_key["upload-ready-pipeline"]["if"]
+    assert 'build.env("WEEKLY") == "1"' in by_key["upload-ready-pipeline"]["if"]
+    assert "NIGHTLY" not in by_key["upload-merge-pipeline"]["if"]
+    assert 'build.branch == "main"' in by_key["upload-merge-pipeline"]["if"]
+    assert 'build.env("WEEKLY") == "1"' in by_key["upload-merge-pipeline"]["if"]
+    assert 'build.env("NIGHTLY") == "1"' in by_key["upload-nightly-pipeline"]["if"]
     assert 'build.env("WEEKLY") == "1"' in rendered
+    assert 'build.env("NON_CRITICAL") == "1"' in rendered
     assert "nightly-test" not in rendered
     assert "weekly-test" not in rendered
     assert "merge-test" not in rendered
@@ -78,6 +86,19 @@ def test_docs_only_allows_main_scheduled_nightly_weekly_only() -> None:
     assert "if: false" not in rendered
 
 
+def test_npu_docs_only_does_not_upload_ready_on_nightly() -> None:
+    """NPU skip_all: scheduled NIGHTLY still uploads L4, not L2 ready."""
+    rendered = _render_bootstrap_pipeline(
+        BOOTSTRAP_STEPS_TEMPLATE,
+        decision=resolve_ci_decision(["docs/foo.md"]),
+        path=Path(".buildkite/npu/bootstrap-upload-steps.yml"),
+    )
+    doc = yaml.safe_load(rendered)
+    by_key = {step["key"]: step for step in doc["steps"]}
+    assert "upload-ready-pipeline" not in by_key
+    assert 'build.env("NIGHTLY") == "1"' in by_key["upload-nightly-pipeline"]["if"]
+
+
 def test_yaml_gated_l45_only_does_not_unconditionally_build_image() -> None:
     rendered = _render([".buildkite/cuda/test-nightly.yml"])
     assert "if: true" not in rendered

tests/dfx/perf/tests/test_qwen3_omni_multi_replicas.json

@@ -10,7 +10,7 @@
           "num_cards": 3
         }
       },
-      "full_model",
+      "slow",
       "omni"
     ],
     "server_params": {

tests/dfx/perf/tests/test_qwen3_omni_vllm_text.json

@@ -10,7 +10,7 @@
           "num_cards": 2
         }
       },
-      "full_model",
+      "slow",
       "omni"
     ],
     "server_params": {

tests/diffusion/quantization/test_quantization_quality.py

@@ -1,5 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
-# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
 """
 Quantization quality gate for diffusion models.
 
@@ -53,6 +53,7 @@ _BENCH_MODULE_NAME = "benchmarks.diffusion.quantization_quality"
 
 if _BENCH_MODULE_NAME not in sys.modules:
     _spec = importlib.util.spec_from_file_location(_BENCH_MODULE_NAME, _BENCH_MODULE_PATH)
+    assert _spec is not None and _spec.loader is not None
     _mod = importlib.util.module_from_spec(_spec)
     sys.modules[_BENCH_MODULE_NAME] = _mod
     _spec.loader.exec_module(_mod)
@@ -93,7 +94,7 @@ class QualityTestConfig:
             return self.quantized_model
         return self.model or ""
 
-    def quantization_ref(self) -> str | None:
+    def quantization_ref(self) -> str | dict[str, object] | None:
         if self.quantized_model is not None:
             return None
         return self.quantization
@@ -500,6 +501,8 @@ def _quality_param(c: QualityTestConfig):
         marks.append(
             pytest.mark.skip(reason="Qwen-Image FP8 quality gate temporarily disabled (see CI / issue tracker).")
         )
+    if c.id == "fp8_ltx2":
+        marks.append(pytest.mark.skip(reason="https://github.com/vllm-project/vllm-omni/issues/6245"))
     return pytest.param(c, id=c.id, marks=marks)
 
 

tests/examples/offline_inference/test_image_task_prompts.py

@@ -3,15 +3,27 @@
 
 from __future__ import annotations
 
+import functools
+import importlib.util
+
 import pytest
 from PIL import Image
 
-from examples.offline_inference.image_to_video.image_to_video import build_image_to_video_prompt
-from examples.offline_inference.text_to_image.text_to_image import build_text_to_image_prompt
+from tests.examples.helpers import EXAMPLES
 
 pytestmark = [pytest.mark.core_model, pytest.mark.cpu]
 
 
+@functools.cache
+def _load_example_module(relative_path: str, module_name: str):
+    path = EXAMPLES / relative_path
+    spec = importlib.util.spec_from_file_location(module_name, path)
+    assert spec is not None and spec.loader is not None
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+    return module
+
+
 @pytest.mark.parametrize(
     ("negative_prompt", "expected_negative_prompt"),
     [(None, None), ("", ""), ("blurry", "blurry")],
@@ -20,7 +32,11 @@ def test_text_to_image_builds_canonical_prompt(
     negative_prompt: str | None,
     expected_negative_prompt: str | None,
 ) -> None:
-    result = build_text_to_image_prompt("a red fox", negative_prompt)
+    mod = _load_example_module(
+        "offline_inference/text_to_image/text_to_image.py",
+        "text_to_image_example",
+    )
+    result = mod.build_text_to_image_prompt("a red fox", negative_prompt)
 
     assert result["prompt"] == "a red fox"
     assert result["modalities"] == ["image"]
@@ -31,9 +47,13 @@ def test_text_to_image_builds_canonical_prompt(
 
 
 def test_image_to_video_builds_canonical_prompt() -> None:
+    mod = _load_example_module(
+        "offline_inference/image_to_video/image_to_video.py",
+        "image_to_video_example",
+    )
     image = Image.new("RGB", (32, 16), "red")
 
-    result = build_image_to_video_prompt(
+    result = mod.build_image_to_video_prompt(
         prompt="the fox turns toward the camera",
         negative_prompt="flicker",
         media_inputs={"image": image},

tests/examples/offline_inference/test_text_to_image.py

@@ -1,3 +1,6 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM-Omni project
+
 """
 Offline inference tests: text-to-image.
 See examples/offline_inference/text_to_image/README.md
@@ -26,6 +29,8 @@ EXAMPLE_OUTPUT_SUBFOLDER = "example_offline_t2i"
 def _skip_readme_snippet(language: str, code: str, h2_title: str) -> tuple[bool, str]:
     if h2_title == "Web UI Demo":
         return True, f"README section '{h2_title}' is intentionally excluded for examples tests"
+    if "krea/Krea-2-Raw" in code:
+        return True, "krea/Krea-2-Raw is gated on Hugging Face Hub; CI has no access"
     return False, ""
 
 

tests/examples/offline_inference/test_text_to_video_prompts.py

@@ -3,19 +3,27 @@
 
 from __future__ import annotations
 
+import functools
+import importlib.util
+
 import pytest
 import torch
 
-from examples.offline_inference.text_to_video.text_to_video import (
-    _MODEL_PRESETS,
-    _detect_preset,
-    _normalize_float_tensor,
-    build_text_to_video_prompt,
-)
+from tests.examples.helpers import EXAMPLES
 
 pytestmark = [pytest.mark.core_model, pytest.mark.cpu]
 
 
+@functools.cache
+def _load_text_to_video():
+    path = EXAMPLES / "offline_inference" / "text_to_video" / "text_to_video.py"
+    spec = importlib.util.spec_from_file_location("text_to_video_example", path)
+    assert spec is not None and spec.loader is not None
+    module = importlib.util.module_from_spec(spec)
+    spec.loader.exec_module(module)
+    return module
+
+
 @pytest.mark.parametrize(
     ("negative_prompt", "expected_negative_prompt"),
     [(None, None), ("", ""), ("blurry", "blurry")],
@@ -24,7 +32,8 @@ def test_text_to_video_builds_canonical_prompt(
     negative_prompt: str | None,
     expected_negative_prompt: str | None,
 ) -> None:
-    result = build_text_to_video_prompt("a robot waves", negative_prompt)
+    mod = _load_text_to_video()
+    result = mod.build_text_to_video_prompt("a robot waves", negative_prompt)
 
     assert result["prompt"] == "a robot waves"
     assert result["modalities"] == ["video"]
@@ -50,11 +59,13 @@ def test_detect_preset_is_scoped_to_model_family(
     model_class_name: str | None,
     preset_name: str,
 ) -> None:
-    assert _detect_preset(model, model_class_name) is _MODEL_PRESETS[preset_name]
+    mod = _load_text_to_video()
+    assert mod._detect_preset(model, model_class_name) is mod._MODEL_PRESETS[preset_name]
 
 
 def test_lingbot_preset_matches_lingbot_defaults() -> None:
-    assert _MODEL_PRESETS["lingbot"] == {
+    mod = _load_text_to_video()
+    assert mod._MODEL_PRESETS["lingbot"] == {
         "model_class_name": "LingBotVideoPipeline",
         "height": 192,
         "width": 320,
@@ -68,24 +79,27 @@ def test_lingbot_preset_matches_lingbot_defaults() -> None:
 
 
 def test_normalize_float_tensor_preserves_zero_to_one_frames() -> None:
+    mod = _load_text_to_video()
     frames = torch.tensor([0.0, 0.25, 1.0, 1.5])
 
-    assert torch.equal(_normalize_float_tensor(frames, "zero_to_one"), torch.tensor([0.0, 0.25, 1.0, 1.0]))
+    assert torch.equal(mod._normalize_float_tensor(frames, "zero_to_one"), torch.tensor([0.0, 0.25, 1.0, 1.0]))
 
 
 def test_normalize_float_tensor_maps_negative_one_to_one_frames() -> None:
+    mod = _load_text_to_video()
     frames = torch.tensor([-1.5, -1.0, 0.0, 1.0, 1.5])
 
     assert torch.equal(
-        _normalize_float_tensor(frames, "negative_one_to_one"),
+        mod._normalize_float_tensor(frames, "negative_one_to_one"),
         torch.tensor([0.0, 0.0, 0.5, 1.0, 1.0]),
     )
 
 
 def test_normalize_float_tensor_uses_one_contract_for_ambiguous_frames() -> None:
+    mod = _load_text_to_video()
     frames = [torch.tensor([0.0, 0.5, 1.0]), torch.tensor([-1.0, 0.0, 1.0])]
 
-    normalized = [_normalize_float_tensor(frame, "negative_one_to_one") for frame in frames]
+    normalized = [mod._normalize_float_tensor(frame, "negative_one_to_one") for frame in frames]
 
     assert torch.equal(normalized[0], torch.tensor([0.5, 0.75, 1.0]))
     assert torch.equal(normalized[1], torch.tensor([0.0, 0.5, 1.0]))

tests/model_executor/models/indextts2/test_optional_dependencies.py

@@ -14,7 +14,7 @@ from packaging.requirements import Requirement
 
 pytestmark = [pytest.mark.core_model, pytest.mark.cpu]
 
-ROOT = Path(__file__).parents[1]
+ROOT = Path(__file__).parents[4]
 TEXT_PROCESSING_PATH = ROOT / "vllm_omni" / "model_executor" / "models" / "indextts2" / "text_processing_v2_5.py"
 INDEXTTS2_FRONTEND_PACKAGES = {
     "fugashi",