oracle-ads 2.13.9rc1__tar.gz → 2.13.10__tar.gz

{oracle_ads-2.13.9rc1 → oracle_ads-2.13.10}/PKG-INFO

@@ -1,8 +1,8 @@
 Metadata-Version: 2.4
 Name: oracle_ads
-Version: 2.13.9rc1
+Version: 2.13.10
 Summary: Oracle Accelerated Data Science SDK
-Keywords: Oracle Cloud Infrastructure,OCI,Machine Learning,ML,Artificial Intelligence,AI,Data Science,Cloud,Oracle,GenAI,Generative AI,Forecast,Anomaly,Document Understanding
+Keywords: Oracle Cloud Infrastructure,OCI,Machine Learning,ML,Artificial Intelligence,AI,Data Science,Cloud,Oracle,GenAI,Generative AI,Forecast,Anomaly,Document Understanding,Anomaly Detection
 Author: Oracle Data Science
 Requires-Python: >=3.8
 Description-Content-Type: text/markdown
@@ -15,7 +15,7 @@ Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
 License-File: LICENSE.txt
-Requires-Dist: PyYAML>=6
+Requires-Dist: PyYAML>=6.0.1
 Requires-Dist: asteval>=0.9.25
 Requires-Dist: cerberus>=1.3.4
 Requires-Dist: cloudpickle>=1.6.0
@@ -99,13 +99,14 @@ Requires-Dist: ipython>=7.23.1, <8.0 ; extra == "notebook"
 Requires-Dist: ipywidgets~=7.6.3 ; extra == "notebook"
 Requires-Dist: lightgbm ; extra == "onnx"
 Requires-Dist: onnx>=1.12.0,<=1.15.0 ; extra == "onnx" and ( python_version < '3.12')
-Requires-Dist: onnx>=1.12.0 ; extra == "onnx" and ( python_version >= '3.12')
-Requires-Dist: onnxmltools>=1.10.0 ; extra == "onnx"
+Requires-Dist: onnx~=1.17.0 ; extra == "onnx" and ( python_version >= '3.12')
+Requires-Dist: onnxmltools~=1.13.0 ; extra == "onnx"
 Requires-Dist: onnxruntime~=1.17.0,!=1.16.0 ; extra == "onnx" and ( python_version < '3.12')
-Requires-Dist: onnxruntime ; extra == "onnx" and ( python_version >= '3.12')
+Requires-Dist: onnxruntime~=1.22.0 ; extra == "onnx" and ( python_version >= '3.12')
 Requires-Dist: oracle_ads[viz] ; extra == "onnx"
 Requires-Dist: protobuf ; extra == "onnx"
-Requires-Dist: skl2onnx>=1.10.4 ; extra == "onnx"
+Requires-Dist: skl2onnx>=1.10.4 ; extra == "onnx" and ( python_version < '3.12')
+Requires-Dist: skl2onnx~=1.18.0 ; extra == "onnx" and ( python_version >= '3.12')
 Requires-Dist: tf2onnx ; extra == "onnx"
 Requires-Dist: xgboost<=1.7 ; extra == "onnx"
 Requires-Dist: conda-pack ; extra == "opctl"

{oracle_ads-2.13.9rc1 → oracle_ads-2.13.10}/ads/aqua/app.py

@@ -64,6 +64,7 @@ class AquaApp:
             set_auth("resource_principal")
         self._auth = default_signer({"service_endpoint": OCI_ODSC_SERVICE_ENDPOINT})
         self.ds_client = oc.OCIClientFactory(**self._auth).data_science
+        self.compute_client = oc.OCIClientFactory(**default_signer()).compute
         self.logging_client = oc.OCIClientFactory(**default_signer()).logging_management
         self.identity_client = oc.OCIClientFactory(**default_signer()).identity
         self.region = extract_region(self._auth)

{oracle_ads-2.13.9rc1 → oracle_ads-2.13.10}/ads/aqua/common/entities.py

@@ -157,6 +157,8 @@ class AquaMultiModelRef(Serializable):
         Optional environment variables to override during deployment.
     artifact_location : Optional[str]
         Artifact path of model in the multimodel group.
+    fine_tune_weights_location : Optional[str]
+        For fine tuned models, the artifact path of the modified model weights
     """
 
     model_id: str = Field(..., description="The model OCID to deploy.")
@@ -171,6 +173,9 @@ class AquaMultiModelRef(Serializable):
     artifact_location: Optional[str] = Field(
         None, description="Artifact path of model in the multimodel group."
     )
+    fine_tune_weights_location: Optional[str] = Field(
+        None, description="For fine tuned models, the artifact path of the modified model weights"
+    )
 
     class Config:
         extra = "ignore"

{oracle_ads-2.13.9rc1 → oracle_ads-2.13.10}/ads/aqua/common/enums.py

@@ -20,6 +20,12 @@ class Resource(ExtendedEnum):
     MODEL_VERSION_SET = "model-version-sets"
 
 
+class PredictEndpoints(ExtendedEnum):
+    CHAT_COMPLETIONS_ENDPOINT = "/v1/chat/completions"
+    TEXT_COMPLETIONS_ENDPOINT = "/v1/completions"
+    EMBEDDING_ENDPOINT = "/v1/embedding"
+
+
 class Tags(ExtendedEnum):
     TASK = "task"
     LICENSE = "license"
@@ -49,6 +55,7 @@ class InferenceContainerType(ExtendedEnum):
 class InferenceContainerTypeFamily(ExtendedEnum):
     AQUA_VLLM_CONTAINER_FAMILY = "odsc-vllm-serving"
     AQUA_VLLM_V1_CONTAINER_FAMILY = "odsc-vllm-serving-v1"
+    AQUA_VLLM_LLAMA4_CONTAINER_FAMILY = "odsc-vllm-serving-llama4"
     AQUA_TGI_CONTAINER_FAMILY = "odsc-tgi-serving"
     AQUA_LLAMA_CPP_CONTAINER_FAMILY = "odsc-llama-cpp-serving"
 
@@ -119,4 +126,9 @@ CONTAINER_FAMILY_COMPATIBILITY: Dict[str, List[str]] = {
         InferenceContainerTypeFamily.AQUA_VLLM_V1_CONTAINER_FAMILY,
         InferenceContainerTypeFamily.AQUA_VLLM_CONTAINER_FAMILY,
     ],
+    InferenceContainerTypeFamily.AQUA_VLLM_LLAMA4_CONTAINER_FAMILY: [
+        InferenceContainerTypeFamily.AQUA_VLLM_LLAMA4_CONTAINER_FAMILY,
+        InferenceContainerTypeFamily.AQUA_VLLM_V1_CONTAINER_FAMILY,
+        InferenceContainerTypeFamily.AQUA_VLLM_CONTAINER_FAMILY,
+    ],
 }

{oracle_ads-2.13.9rc1 → oracle_ads-2.13.10}/ads/aqua/common/utils.py

@@ -832,7 +832,9 @@ def get_params_dict(params: Union[str, List[str]]) -> dict:
     """
     params_list = get_params_list(params) if isinstance(params, str) else params
     return {
-        split_result[0]: split_result[1] if len(split_result) > 1 else UNKNOWN
+        split_result[0]: " ".join(split_result[1:])
+        if len(split_result) > 1
+        else UNKNOWN
         for split_result in (x.split() for x in params_list)
     }
 
@@ -881,7 +883,9 @@ def build_params_string(params: dict) -> str:
         A params string.
     """
     return (
-        " ".join(f"{name} {value}" for name, value in params.items()).strip()
+        " ".join(
+            f"{name} {value}" if value else f"{name}" for name, value in params.items()
+        ).strip()
         if params
         else UNKNOWN
     )
@@ -1158,9 +1162,11 @@ def validate_cmd_var(cmd_var: List[str], overrides: List[str]) -> List[str]:
 
 
 def build_pydantic_error_message(ex: ValidationError):
-    """Added to handle error messages from pydantic model validator.
+    """
+    Added to handle error messages from pydantic model validator.
     Combine both loc and msg for errors where loc (field) is present in error details, else only build error
-    message using msg field."""
+    message using msg field.
+    """
 
     return {
         ".".join(map(str, e["loc"])): e["msg"]
@@ -1185,67 +1191,71 @@ def is_pydantic_model(obj: object) -> bool:
 
 @cached(cache=TTLCache(maxsize=1, ttl=timedelta(minutes=5), timer=datetime.now))
 def load_gpu_shapes_index(
-    auth: Optional[Dict] = None,
+    auth: Optional[Dict[str, Any]] = None,
 ) -> GPUShapesIndex:
     """
-    Loads the GPU shapes index from Object Storage or a local resource folder.
+    Load the GPU shapes index, preferring the OS bucket copy over the local one.
 
-    The function first attempts to load the file from an Object Storage bucket using fsspec.
-    If the loading fails (due to connection issues, missing file, etc.), it falls back to
-    loading the index from a local file.
+    Attempts to read `gpu_shapes_index.json` from OCI Object Storage first;
+    if that succeeds, those entries will override the local defaults.
 
     Parameters
     ----------
-    auth: (Dict, optional). Defaults to None.
-        The default authentication 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
-        authentication signer and kwargs required to instantiate IdentityClient object.
+    auth
+        Optional auth dict (as returned by `ads.common.auth.default_signer()`)
+        to pass through to `fsspec.open()`.
 
     Returns
     -------
-    GPUShapesIndex: The parsed GPU shapes index.
+    GPUShapesIndex
+        Merged index where any shape present remotely supersedes the local entry.
 
     Raises
     ------
-    FileNotFoundError: If the GPU shapes index cannot be found in either Object Storage or locally.
-    json.JSONDecodeError: If the JSON is malformed.
+    json.JSONDecodeError
+        If any of the JSON is malformed.
     """
     file_name = "gpu_shapes_index.json"
-    data: Dict[str, Any] = {}
 
-    # Check if the CONDA_BUCKET_NS environment variable is set.
+    # Try remote load
+    remote_data: Dict[str, Any] = {}
     if CONDA_BUCKET_NS:
         try:
             auth = auth or authutil.default_signer()
-            # Construct the object storage path. Adjust bucket name and path as needed.
             storage_path = (
                 f"oci://{CONDA_BUCKET_NAME}@{CONDA_BUCKET_NS}/service_pack/{file_name}"
             )
-            logger.debug("Loading GPU shapes index from Object Storage")
-            with fsspec.open(storage_path, mode="r", **auth) as file_obj:
-                data = json.load(file_obj)
-            logger.debug("Successfully loaded GPU shapes index.")
-        except Exception as ex:
             logger.debug(
-                f"Failed to load GPU shapes index from Object Storage. Details: {ex}"
-            )
-
-    # If loading from Object Storage failed, load from the local resource folder.
-    if not data:
-        try:
-            local_path = os.path.join(
-                os.path.dirname(__file__), "../resources", file_name
+                "Loading GPU shapes index from Object Storage: %s", storage_path
             )
-            logger.debug(f"Loading GPU shapes index from {local_path}.")
-            with open(local_path) as file_obj:
-                data = json.load(file_obj)
-            logger.debug("Successfully loaded GPU shapes index.")
-        except Exception as e:
+            with fsspec.open(storage_path, mode="r", **auth) as f:
+                remote_data = json.load(f)
             logger.debug(
-                f"Failed to load GPU shapes index from {local_path}. Details: {e}"
+                "Loaded %d shapes from Object Storage",
+                len(remote_data.get("shapes", {})),
             )
+        except Exception as ex:
+            logger.debug("Remote load failed (%s); falling back to local", ex)
+
+    # Load local copy
+    local_data: Dict[str, Any] = {}
+    local_path = os.path.join(os.path.dirname(__file__), "../resources", file_name)
+    try:
+        logger.debug("Loading GPU shapes index from local file: %s", local_path)
+        with open(local_path) as f:
+            local_data = json.load(f)
+        logger.debug(
+            "Loaded %d shapes from local file", len(local_data.get("shapes", {}))
+        )
+    except Exception as ex:
+        logger.debug("Local load GPU shapes index failed (%s)", ex)
+
+    # Merge: remote shapes override local
+    local_shapes = local_data.get("shapes", {})
+    remote_shapes = remote_data.get("shapes", {})
+    merged_shapes = {**local_shapes, **remote_shapes}
 
-    return GPUShapesIndex(**data)
+    return GPUShapesIndex(shapes=merged_shapes)
 
 
 def get_preferred_compatible_family(selected_families: set[str]) -> str:

{oracle_ads-2.13.9rc1 → oracle_ads-2.13.10}/ads/aqua/config/container_config.py

@@ -7,6 +7,7 @@ from typing import Dict, List, Optional
 from oci.data_science.models import ContainerSummary
 from pydantic import Field
 
+from ads.aqua import logger
 from ads.aqua.config.utils.serializer import Serializable
 from ads.aqua.constants import (
     SERVICE_MANAGED_CONTAINER_URI_SCHEME,
@@ -168,50 +169,46 @@ class AquaContainerConfig(Serializable):
             container_type = container.family_name
             usages = [x.upper() for x in container.usages]
             if "INFERENCE" in usages or "MULTI_MODEL" in usages:
+                # Extract additional configurations
+                additional_configurations = {}
+                try:
+                    additional_configurations = (
+                        container.workload_configuration_details_list[
+                            0
+                        ].additional_configurations
+                    )
+                except (AttributeError, IndexError) as ex:
+                    logger.debug(
+                        "Failed to extract `additional_configurations` for container '%s': %s",
+                        getattr(container, "container_name", "<unknown>"),
+                        ex,
+                    )
+
                 container_item.platforms.append(
-                    container.workload_configuration_details_list[
-                        0
-                    ].additional_configurations.get("platforms")
+                    additional_configurations.get("platforms")
                 )
                 container_item.model_formats.append(
-                    container.workload_configuration_details_list[
-                        0
-                    ].additional_configurations.get("modelFormats")
+                    additional_configurations.get("modelFormats")
                 )
+
+                # Parse environment variables from `additional_configurations`.
+                # Only keys present in the configuration will be added to the result.
+                config_keys = {
+                    "MODEL_DEPLOY_PREDICT_ENDPOINT": UNKNOWN,
+                    "MODEL_DEPLOY_HEALTH_ENDPOINT": UNKNOWN,
+                    "PORT": UNKNOWN,
+                    "HEALTH_CHECK_PORT": UNKNOWN,
+                    "VLLM_USE_V1": UNKNOWN,
+                }
+
                 env_vars = [
-                    {
-                        "MODEL_DEPLOY_PREDICT_ENDPOINT": container.workload_configuration_details_list[
-                            0
-                        ].additional_configurations.get(
-                            "MODEL_DEPLOY_PREDICT_ENDPOINT", UNKNOWN
-                        )
-                    },
-                    {
-                        "MODEL_DEPLOY_HEALTH_ENDPOINT": container.workload_configuration_details_list[
-                            0
-                        ].additional_configurations.get(
-                            "MODEL_DEPLOY_HEALTH_ENDPOINT", UNKNOWN
-                        )
-                    },
-                    {
-                        "MODEL_DEPLOY_ENABLE_STREAMING": container.workload_configuration_details_list[
-                            0
-                        ].additional_configurations.get(
-                            "MODEL_DEPLOY_ENABLE_STREAMING", UNKNOWN
-                        )
-                    },
-                    {
-                        "PORT": container.workload_configuration_details_list[
-                            0
-                        ].additional_configurations.get("PORT", "")
-                    },
-                    {
-                        "HEALTH_CHECK_PORT": container.workload_configuration_details_list[
-                            0
-                        ].additional_configurations.get("HEALTH_CHECK_PORT", UNKNOWN),
-                    },
+                    {key: additional_configurations.get(key, default)}
+                    for key, default in config_keys.items()
+                    if key in additional_configurations
                 ]
-                container_spec = AquaContainerConfigSpec(
+
+                # Build container spec
+                container_item.spec = AquaContainerConfigSpec(
                     cli_param=container.workload_configuration_details_list[0].cmd,
                     server_port=str(
                         container.workload_configuration_details_list[0].server_port
@@ -236,13 +233,14 @@ class AquaContainerConfig(Serializable):
                         )
                     ),
                 )
-                container_item.spec = container_spec
+
             if "INFERENCE" in usages or "MULTI_MODEL" in usages:
                 inference_items[container_type] = container_item
             if "FINE_TUNE" in usages:
                 finetune_items[container_type] = container_item
             if "EVALUATION" in usages:
                 evaluate_items[container_type] = container_item
+
         return cls(
             inference=inference_items, finetune=finetune_items, evaluate=evaluate_items
         )

{oracle_ads-2.13.9rc1 → oracle_ads-2.13.10}/ads/aqua/extension/deployment_handler.py

@@ -2,16 +2,18 @@
 # Copyright (c) 2024, 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 typing import List, Union
+from typing import List, Optional, Union
 from urllib.parse import urlparse
 
 from tornado.web import HTTPError
 
+from ads.aqua.app import logger
+from ads.aqua.client.client import Client, ExtendedRequestError
 from ads.aqua.common.decorator import handle_exceptions
+from ads.aqua.common.enums import PredictEndpoints
 from ads.aqua.extension.base_handler import AquaAPIhandler
 from ads.aqua.extension.errors import Errors
-from ads.aqua.modeldeployment import AquaDeploymentApp, MDInferenceResponse
-from ads.aqua.modeldeployment.entities import ModelParams
+from ads.aqua.modeldeployment import AquaDeploymentApp
 from ads.config import COMPARTMENT_OCID
 
 
@@ -175,23 +177,107 @@ class AquaDeploymentHandler(AquaAPIhandler):
         )
 
 
-class AquaDeploymentInferenceHandler(AquaAPIhandler):
-    @staticmethod
-    def validate_predict_url(endpoint):
-        try:
-            url = urlparse(endpoint)
-            if url.scheme != "https":
-                return False
-            if not url.netloc:
-                return False
-            return url.path.endswith("/predict")
-        except Exception:
-            return False
+class AquaDeploymentStreamingInferenceHandler(AquaAPIhandler):
+    def _get_model_deployment_response(
+        self,
+        model_deployment_id: str,
+        payload: dict,
+        route_override_header: Optional[str],
+    ):
+        """
+        Returns the model deployment inference response in a streaming fashion.
+
+        This method connects to the specified model deployment endpoint and
+        streams the inference output back to the caller, handling both text
+        and chat completion endpoints depending on the route override.
+
+        Parameters
+        ----------
+        model_deployment_id : str
+            The OCID of the model deployment to invoke.
+            Example: 'ocid1.datasciencemodeldeployment.iad.oc1.xxxyz'
+
+        payload : dict
+            Dictionary containing the model inference parameters.
+            Same example for text completions:
+                {
+                    "max_tokens": 1024,
+                    "temperature": 0.5,
+                    "prompt": "what are some good skills deep learning expert. Give us some tips on how to structure interview with some coding example?",
+                    "top_p": 0.4,
+                    "top_k": 100,
+                    "model": "odsc-llm",
+                    "frequency_penalty": 1,
+                    "presence_penalty": 1,
+                    "stream": true
+                }
+
+        route_override_header : Optional[str]
+            Optional override for the inference route, used for routing between
+            different endpoint types (e.g., chat vs. text completions).
+            Example: '/v1/chat/completions'
+
+        Returns
+        -------
+        Generator[str]
+            A generator that yields strings of the model's output as they are received.
+
+        Raises
+        ------
+        HTTPError
+            If the request to the model deployment fails or if streaming cannot be established.
+        """
+
+        model_deployment = AquaDeploymentApp().get(model_deployment_id)
+        endpoint = model_deployment.endpoint + "/predictWithResponseStream"
+        endpoint_type = model_deployment.environment_variables.get(
+            "MODEL_DEPLOY_PREDICT_ENDPOINT", PredictEndpoints.TEXT_COMPLETIONS_ENDPOINT
+        )
+        aqua_client = Client(endpoint=endpoint)
+
+        if PredictEndpoints.CHAT_COMPLETIONS_ENDPOINT in (
+            endpoint_type,
+            route_override_header,
+        ):
+            try:
+                for chunk in aqua_client.chat(
+                    messages=payload.pop("messages"),
+                    payload=payload,
+                    stream=True,
+                ):
+                    try:
+                        yield chunk["choices"][0]["delta"]["content"]
+                    except Exception as e:
+                        logger.debug(
+                            f"Exception occurred while parsing streaming response: {e}"
+                        )
+            except ExtendedRequestError as ex:
+                raise HTTPError(400, str(ex))
+            except Exception as ex:
+                raise HTTPError(500, str(ex))
+
+        elif endpoint_type == PredictEndpoints.TEXT_COMPLETIONS_ENDPOINT:
+            try:
+                for chunk in aqua_client.generate(
+                    prompt=payload.pop("prompt"),
+                    payload=payload,
+                    stream=True,
+                ):
+                    try:
+                        yield chunk["choices"][0]["text"]
+                    except Exception as e:
+                        logger.debug(
+                            f"Exception occurred while parsing streaming response: {e}"
+                        )
+            except ExtendedRequestError as ex:
+                raise HTTPError(400, str(ex))
+            except Exception as ex:
+                raise HTTPError(500, str(ex))
 
     @handle_exceptions
-    def post(self, *args, **kwargs):  # noqa: ARG002
+    def post(self, model_deployment_id):
         """
-        Handles inference request for the Active Model Deployments
+        Handles streaming inference request for the Active Model Deployments
         Raises
         ------
         HTTPError
@@ -205,32 +291,29 @@ class AquaDeploymentInferenceHandler(AquaAPIhandler):
         if not input_data:
             raise HTTPError(400, Errors.NO_INPUT_DATA)
 
-        endpoint = input_data.get("endpoint")
-        if not endpoint:
-            raise HTTPError(400, Errors.MISSING_REQUIRED_PARAMETER.format("endpoint"))
-
-        if not self.validate_predict_url(endpoint):
-            raise HTTPError(400, Errors.INVALID_INPUT_DATA_FORMAT.format("endpoint"))
-
         prompt = input_data.get("prompt")
-        if not prompt:
-            raise HTTPError(400, Errors.MISSING_REQUIRED_PARAMETER.format("prompt"))
+        messages = input_data.get("messages")
 
-        model_params = (
-            input_data.get("model_params") if input_data.get("model_params") else {}
-        )
-        try:
-            model_params_obj = ModelParams(**model_params)
-        except Exception as ex:
+        if not prompt and not messages:
             raise HTTPError(
-                400, Errors.INVALID_INPUT_DATA_FORMAT.format("model_params")
-            ) from ex
-
-        return self.finish(
-            MDInferenceResponse(prompt, model_params_obj).get_model_deployment_response(
-                endpoint
+                400, Errors.MISSING_REQUIRED_PARAMETER.format("prompt/messages")
             )
+        if not input_data.get("model"):
+            raise HTTPError(400, Errors.MISSING_REQUIRED_PARAMETER.format("model"))
+        route_override_header = self.request.headers.get("route", None)
+        self.set_header("Content-Type", "text/event-stream")
+        response_gen = self._get_model_deployment_response(
+            model_deployment_id, input_data, route_override_header
         )
+        try:
+            for chunk in response_gen:
+                self.write(chunk)
+                self.flush()
+            self.finish()
+        except Exception as ex:
+            self.set_status(ex.status_code)
+            self.write({"message": "Error occurred", "reason": str(ex)})
+            self.finish()
 
 
 class AquaDeploymentParamsHandler(AquaAPIhandler):
@@ -294,5 +377,5 @@ __handlers__ = [
     ("deployments/?([^/]*)", AquaDeploymentHandler),
     ("deployments/?([^/]*)/activate", AquaDeploymentHandler),
     ("deployments/?([^/]*)/deactivate", AquaDeploymentHandler),
-    ("inference", AquaDeploymentInferenceHandler),
+    ("inference/stream/?([^/]*)", AquaDeploymentStreamingInferenceHandler),
 ]

{oracle_ads-2.13.9rc1 → oracle_ads-2.13.10}/ads/aqua/extension/ui_handler.py

@@ -15,7 +15,7 @@ from ads.aqua.extension.errors import Errors
 from ads.aqua.extension.utils import validate_function_parameters
 from ads.aqua.model.entities import ImportModelDetails
 from ads.aqua.ui import AquaUIApp
-from ads.config import COMPARTMENT_OCID
+from ads.config import COMPARTMENT_OCID, IS_BYOR_ENABLED
 
 
 @dataclass
@@ -82,6 +82,10 @@ class AquaUIHandler(AquaAPIhandler):
             return self.is_bucket_versioned()
         elif paths.startswith("aqua/containers"):
             return self.list_containers()
+        elif paths.startswith("aqua/capacityreservations/enabled"):
+            return self.is_capacity_reservations_enabled()
+        elif paths.startswith("aqua/capacityreservations"):
+            return self.list_capacity_reservations()
         else:
             raise HTTPError(400, f"The request {self.request.path} is invalid.")
 
@@ -103,6 +107,19 @@ class AquaUIHandler(AquaAPIhandler):
             AquaUIApp().list_log_groups(compartment_id=compartment_id, **kwargs)
         )
 
+    def is_capacity_reservations_enabled(self):
+        """Checks if the tenant is whitelisted for BYOR (Bring your own reservation) feature."""
+        return self.finish({"status": str(IS_BYOR_ENABLED).strip().lower() == "true"})
+
+    def list_capacity_reservations(self, **kwargs):
+        """Lists users compute reservations in a specified compartment."""
+        compartment_id = self.get_argument("compartment_id", default=COMPARTMENT_OCID)
+        return self.finish(
+            AquaUIApp().list_capacity_reservations(
+                compartment_id=compartment_id, **kwargs
+            )
+        )
+
     def list_logs(self, log_group_id: str, **kwargs):
         """Lists the specified log group's log objects."""
         return self.finish(AquaUIApp().list_logs(log_group_id=log_group_id, **kwargs))
@@ -279,4 +296,5 @@ __handlers__ = [
     ("bucket/versioning/?([^/]*)", AquaUIHandler),
     ("containers/?([^/]*)", AquaUIHandler),
     ("cli/?([^/]*)", AquaCLIHandler),
+    ("capacityreservations/?([^/]*)", AquaUIHandler),
 ]

{oracle_ads-2.13.9rc1 → oracle_ads-2.13.10}/ads/aqua/model/model.py

@@ -83,6 +83,10 @@ from ads.aqua.model.entities import (
     ModelValidationResult,
 )
 from ads.aqua.model.enums import MultiModelSupportedTaskType
+from ads.aqua.model.utils import (
+    extract_base_model_from_ft,
+    extract_fine_tune_artifacts_path,
+)
 from ads.common.auth import default_signer
 from ads.common.oci_resource import SEARCH_TYPE, OCIResource
 from ads.common.utils import (
@@ -311,12 +315,27 @@ class AquaModelApp(AquaApp):
             #         "Currently only service models are supported for multi model deployment."
             #     )
 
+            # check if model is a fine-tuned model and if so, add the fine tuned weights path to the fine_tune_weights_location pydantic field
+            is_fine_tuned_model = (
+                Tags.AQUA_FINE_TUNED_MODEL_TAG in source_model.freeform_tags
+            )
+
+            if is_fine_tuned_model:
+                model.model_id, model.model_name = extract_base_model_from_ft(
+                    source_model
+                )
+                model_artifact_path, model.fine_tune_weights_location = (
+                    extract_fine_tune_artifacts_path(source_model)
+                )
+
+            else:
+                # Retrieve model artifact for base models
+                model_artifact_path = source_model.artifact
+
             display_name_list.append(display_name)
 
             self._extract_model_task(model, source_model)
 
-            # Retrieve model artifact
-            model_artifact_path = source_model.artifact
             if not model_artifact_path:
                 raise AquaValueError(
                     f"Model '{display_name}' (ID: {model.model_id}) has no artifacts. "
@@ -367,7 +386,8 @@ class AquaModelApp(AquaApp):
                 raise AquaValueError(
                     "The selected models are associated with different container families: "
                     f"{list(selected_models_deployment_containers)}."
-                    "For multi-model deployment, all models in the group must share the same container family."
+                    "For multi-model deployment, all models in the group must belong to the same container "
+                    "family or to compatible container families."
                 )
         else:
             deployment_container = selected_models_deployment_containers.pop()