Added artifacts validation.

Author: lu-ohai

ads/model/__init__.py

@@ -1,6 +1,6 @@
 #!/usr/bin/env python
 
-# Copyright (c) 2021, 2024 Oracle and/or its affiliates.
+# Copyright (c) 2021, 2025 Oracle and/or its affiliates.
 # Licensed under the Universal Permissive License v 1.0 as shown at https://oss.oracle.com/licenses/upl/
 
 from ads.model.datascience_model import DataScienceModel

ads/model/artifact.py

@@ -1,6 +1,6 @@
 #!/usr/bin/env python
 
-# Copyright (c) 2022, 2024 Oracle and/or its affiliates.
+# Copyright (c) 2022, 2025 Oracle and/or its affiliates.
 # Licensed under the Universal Permissive License v 1.0 as shown at https://oss.oracle.com/licenses/upl/
 
 import fnmatch

ads/model/extractor/embedding_onnx_extractor.py

@@ -1,6 +1,6 @@
 #!/usr/bin/env python
 
-# Copyright (c) 2024 Oracle and/or its affiliates.
+# Copyright (c) 2025 Oracle and/or its affiliates.
 # Licensed under the Universal Permissive License v 1.0 as shown at https://oss.oracle.com/licenses/upl/
 
 from ads.common.decorator.runtime_dependency import (

ads/model/framework/embedding_onnx_model.py

@@ -1,13 +1,27 @@
 #!/usr/bin/env python
 
-# Copyright (c) 2024 Oracle and/or its affiliates.
+# Copyright (c) 2025 Oracle and/or its affiliates.
 # Licensed under the Universal Permissive License v 1.0 as shown at https://oss.oracle.com/licenses/upl/
 
+import logging
+import os
+from pathlib import Path
 from typing import Dict, Optional
 
 from ads.model.extractor.embedding_onnx_extractor import EmbeddingONNXExtractor
 from ads.model.generic_model import FrameworkSpecificModel
 
+logger = logging.getLogger(__name__)
+
+CONFIG = "config.json"
+TOKENIZERS = [
+    "tokenizer.json",
+    "tokenizer_config.json",
+    "spiece.model",
+    "vocab.txt",
+    "vocab.json",
+]
+
 
 class EmbeddingONNXModel(FrameworkSpecificModel):
     """EmbeddingONNXModel class for embedding onnx model.
@@ -18,6 +32,12 @@ class EmbeddingONNXModel(FrameworkSpecificModel):
         The algorithm of the model.
     artifact_dir: str
         Artifact directory to store the files needed for deployment.
+    model_file_name: str
+        Path to the model artifact.
+    config_json: str
+        Path to the config.json file.
+    tokenizer_dir: str
+        Path to the tokenizer directory.
     auth: Dict
         Default authentication is set using the `ads.set_auth` API. To override the
         default, use the `ads.common.auth.api_keys` or `ads.common.auth.resource_principal` to create
@@ -166,6 +186,9 @@ class EmbeddingONNXModel(FrameworkSpecificModel):
     def __init__(
         self,
         artifact_dir: Optional[str] = None,
+        model_file_name: Optional[str] = None,
+        config_json: Optional[str] = None,
+        tokenizer_dir: Optional[str] = None,
         auth: Optional[Dict] = None,
         serialize: bool = False,
         **kwargs: dict,
@@ -175,8 +198,14 @@ def __init__(
 
         Parameters
         ----------
-        artifact_dir: str
+        artifact_dir: (str, optional). Defaults to None.
             Directory for generate artifact.
+        model_file_name: (str, optional). Defaults to None.
+            Path to the model artifact.
+        config_json: (str, optional). Defaults to None.
+            Path to the config.json file.
+        tokenizer_dir: (str, optional). Defaults to None.
+            Path to the tokenizer directory.
         auth: (Dict, optional). Defaults to None.
             The default authetication is set using `ads.set_auth` API. If you need to override the
             default, use the `ads.common.auth.api_keys` or `ads.common.auth.resource_principal` to create appropriate
@@ -260,12 +289,63 @@ def __init__(
             **kwargs,
         )
 
+        self._validate_artifact_directory(
+            model_file_name=model_file_name,
+            config_json=config_json,
+            tokenizer_dir=tokenizer_dir,
+        )
+
         self._extractor = EmbeddingONNXExtractor()
         self.framework = self._extractor.framework
         self.algorithm = self._extractor.algorithm
         self.version = self._extractor.version
         self.hyperparameter = self._extractor.hyperparameter
 
+    def _validate_artifact_directory(
+        self,
+        model_file_name: str = None,
+        config_json: str = None,
+        tokenizer_dir: str = None,
+    ):
+        artifacts = []
+        for _, _, files in os.walk(self.artifact_dir):
+            artifacts.extend(files)
+
+        if not artifacts:
+            raise ValueError(
+                f"No files found in {self.artifact_dir}. Specify a valid `artifact_dir`."
+            )
+
+        if not model_file_name:
+            has_model_file = False
+            for artifact in artifacts:
+                if Path(artifact).suffix.lstrip(".").lower() == "onnx":
+                    has_model_file = True
+                    break
+
+            if not has_model_file:
+                raise ValueError(
+                    f"No onnx model found in {self.artifact_dir}. Specify a valid `artifact_dir` or `model_file_name`."
+                )
+
+        if not config_json:
+            if CONFIG not in artifacts:
+                logger.warning(
+                    f"No {CONFIG} found in {self.artifact_dir}. Specify a valid `artifact_dir` or `config_json`."
+                )
+
+        if not tokenizer_dir:
+            has_tokenizer = False
+            for artifact in artifacts:
+                if artifact in TOKENIZERS:
+                    has_tokenizer = True
+                    break
+
+            if not has_tokenizer:
+                logger.warning(
+                    f"No tokenizer found in {self.artifact_dir}. Specify a valid `artifact_dir` or `tokenizer_dir`."
+                )
+
     def verify(
         self, data=None, reload_artifacts=True, auto_serialize_data=False, **kwargs
     ):

tests/unitary/with_extras/model/test_model_framework_embedding_onnx_model.py

@@ -1,6 +1,6 @@
 #!/usr/bin/env python
 
-# Copyright (c) 2024 Oracle and/or its affiliates.
+# Copyright (c) 2025 Oracle and/or its affiliates.
 # Licensed under the Universal Permissive License v 1.0 as shown at https://oss.oracle.com/licenses/upl/
 
 import os
@@ -21,13 +21,20 @@ def setup_class(cls):
         cls.inference_conda = "oci://fake_bucket@fake_namespace/inference_conda"
         cls.training_conda = "oci://fake_bucket@fake_namespace/training_conda"
 
-    def test_init(self):
+    @patch(
+        "ads.model.framework.embedding_onnx_model.EmbeddingONNXModel._validate_artifact_directory"
+    )
+    def test_init(self, mock_validate):
         model = EmbeddingONNXModel(artifact_dir=self.tmp_model_dir)
         assert model.algorithm == "Embedding_ONNX"
         assert model.framework == Framework.EMBEDDING_ONNX
+        mock_validate.assert_called()
 
     @patch("ads.model.generic_model.GenericModel.verify")
-    def test_prepare_and_verify(self, mock_verify):
+    @patch(
+        "ads.model.framework.embedding_onnx_model.EmbeddingONNXModel._validate_artifact_directory"
+    )
+    def test_prepare_and_verify(self, mock_validate, mock_verify):
         mock_verify.return_value = {"results": "successful"}
 
         model = EmbeddingONNXModel(artifact_dir=self.tmp_model_dir)
@@ -87,11 +94,17 @@ def test_prepare_and_verify(self, mock_verify):
             reload_artifacts=True,
             auto_serialize_data=False,
         )
+        mock_validate.assert_called()
 
     @patch("ads.model.generic_model.GenericModel.predict")
     @patch("ads.model.generic_model.GenericModel.deploy")
     @patch("ads.model.generic_model.GenericModel.save")
-    def test_prepare_save_deploy_predict(self, mock_save, mock_deploy, mock_predict):
+    @patch(
+        "ads.model.framework.embedding_onnx_model.EmbeddingONNXModel._validate_artifact_directory"
+    )
+    def test_prepare_save_deploy_predict(
+        self, mock_validate, mock_save, mock_deploy, mock_predict
+    ):
         model = EmbeddingONNXModel(artifact_dir=self.tmp_model_dir)
         model.prepare(
             model_file_name="test_model_file_name",
@@ -127,6 +140,7 @@ def test_prepare_save_deploy_predict(self, mock_save, mock_deploy, mock_predict)
             deployment_ocpus=20,
             deployment_memory_in_gbs=256,
         )
+        mock_validate.assert_called()
 
     def teardown_class(cls):
         shutil.rmtree(cls.tmp_model_dir, ignore_errors=True)