[CI/Build] Fix v1 OOT registration test (#25547)

Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>

tests/conftest.py

@@ -1079,7 +1079,7 @@ def dummy_llava_path():
                           local_dir=_dummy_llava_path,
                           ignore_patterns=[
                               "*.bin", "*.bin.index.json", "*.pt", "*.h5",
-                              "*.msgpack"
+                              "*.msgpack", "*.safetensors"
                           ])
         assert os.path.exists(json_path)
         with open(json_path) as f:
@@ -1098,7 +1098,7 @@ def dummy_gemma2_embedding_path():
                           local_dir=_dummy_gemma2_embedding_path,
                           ignore_patterns=[
                               "*.bin", "*.bin.index.json", "*.pt", "*.h5",
-                              "*.msgpack"
+                              "*.msgpack", "*.safetensors"
                           ])
         assert os.path.exists(json_path)
         with open(json_path) as f:

tests/models/test_oot_registration.py

@@ -42,7 +42,6 @@ def test_oot_registration_text_generation(
             assert rest == ""
 
 
-@pytest.mark.skip(reason="This test is skipped because it failed on V1.")
 @create_new_process_for_each_test()
 def test_oot_registration_embedding(
     monkeypatch: pytest.MonkeyPatch,
@@ -63,7 +62,6 @@ def test_oot_registration_embedding(
 image = convert_image_mode(ImageAsset("cherry_blossom").pil_image, "RGB")
 
 
-@pytest.mark.skip(reason="This test is skipped because it failed on V1.")
 @create_new_process_for_each_test()
 def test_oot_registration_multimodal(
     monkeypatch: pytest.MonkeyPatch,

vllm/model_executor/models/registry.py

@@ -483,23 +483,23 @@ class _LazyRegisteredModel(_BaseRegisteredModel):
     def inspect_model_cls(self) -> _ModelInfo:
         model_path = Path(
             __file__).parent / f"{self.module_name.split('.')[-1]}.py"
+        module_hash = None
 
-        assert model_path.exists(), \
-            f"Model {self.module_name} expected to be on path {model_path}"
-        with open(model_path, "rb") as f:
-            module_hash = hashlib.md5(f.read()).hexdigest()
-
-        mi = self._load_modelinfo_from_cache(module_hash)
-        if mi is not None:
-            logger.debug(("Loaded model info "
-                          "for class %s.%s from cache"), self.module_name,
-                         self.class_name)
-            return mi
-        else:
-            logger.debug(("Cache model info "
-                          "for class %s.%s miss. "
-                          "Loading model instead."), self.module_name,
-                         self.class_name)
+        if model_path.exists():
+            with open(model_path, "rb") as f:
+                module_hash = hashlib.md5(f.read()).hexdigest()
+
+            mi = self._load_modelinfo_from_cache(module_hash)
+            if mi is not None:
+                logger.debug(("Loaded model info "
+                              "for class %s.%s from cache"), self.module_name,
+                             self.class_name)
+                return mi
+            else:
+                logger.debug(("Cache model info "
+                              "for class %s.%s miss. "
+                              "Loading model instead."), self.module_name,
+                             self.class_name)
 
         # Performed in another process to avoid initializing CUDA
         mi = _run_in_subprocess(
@@ -508,7 +508,8 @@ class _LazyRegisteredModel(_BaseRegisteredModel):
                      self.class_name)
 
         # save cache file
-        self._save_modelinfo_to_cache(mi, module_hash)
+        if module_hash is not None:
+            self._save_modelinfo_to_cache(mi, module_hash)
 
         return mi