diff --git a/tests/conftest.py b/tests/conftest.py index dc70c98359598..a50985a465e6c 100644 --- a/tests/conftest.py +++ b/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: diff --git a/tests/models/test_oot_registration.py b/tests/models/test_oot_registration.py index 9b376f2a260ac..4aa7bb7297893 100644 --- a/tests/models/test_oot_registration.py +++ b/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, diff --git a/vllm/model_executor/models/registry.py b/vllm/model_executor/models/registry.py index 6ab3fa902c387..ac0ec6ca146c9 100644 --- a/vllm/model_executor/models/registry.py +++ b/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() + 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) + 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