Updates to notebook; use direct requests to NeMo where needed

docs/_static/llama-stack-spec.html

@@ -6347,7 +6347,36 @@
                         "default": "model"
                     },
                     "model": {
-                        "type": "string",
+                        "oneOf": [
+                            {
+                                "type": "string"
+                            },
+                            {
+                                "type": "object",
+                                "additionalProperties": {
+                                    "oneOf": [
+                                        {
+                                            "type": "null"
+                                        },
+                                        {
+                                            "type": "boolean"
+                                        },
+                                        {
+                                            "type": "number"
+                                        },
+                                        {
+                                            "type": "string"
+                                        },
+                                        {
+                                            "type": "array"
+                                        },
+                                        {
+                                            "type": "object"
+                                        }
+                                    ]
+                                }
+                            }
+                        ],
                         "description": "The model ID to evaluate."
                     },
                     "sampling_params": {
@@ -6362,8 +6391,7 @@
                 "additionalProperties": false,
                 "required": [
                     "type",
-                    "model",
+                    "model"
-                    "sampling_params"
                 ],
                 "title": "ModelCandidate",
                 "description": "A model candidate for evaluation."

docs/_static/llama-stack-spec.yaml

@@ -4468,7 +4468,17 @@ components:
           const: model
           default: model
         model:
-          type: string
+          oneOf:
+            - type: string
+            - type: object
+              additionalProperties:
+                oneOf:
+                  - type: 'null'
+                  - type: boolean
+                  - type: number
+                  - type: string
+                  - type: array
+                  - type: object
           description: The model ID to evaluate.
         sampling_params:
           $ref: '#/components/schemas/SamplingParams'
@@ -4482,7 +4492,6 @@ components:
       required:
         - type
         - model
-        - sampling_params
       title: ModelCandidate
       description: A model candidate for evaluation.
     RegexParserScoringFnParams:

llama_stack/apis/eval/eval.py

@@ -27,8 +27,8 @@ class ModelCandidate(BaseModel):
     """
 
     type: Literal["model"] = "model"
-    model: str
+    model: Union[str, Dict[str, Any]]
-    sampling_params: SamplingParams
+    sampling_params: Optional[SamplingParams] = Field(default_factory=SamplingParams)
     system_message: Optional[SystemMessage] = None
 
 

llama_stack/providers/remote/eval/nvidia/eval.py

@@ -48,13 +48,13 @@ class NVIDIAEvalImpl(
 
     async def _evaluator_get(self, path):
         """Helper for making GET requests to the evaluator service."""
-        response = requests.get(url=f"{self.config.evaluator_service_url}/{path}")
+        response = requests.get(url=f"{self.config.evaluator_service_url}{path}")
         response.raise_for_status()
         return response.json()
 
     async def _evaluator_post(self, path, data):
         """Helper for making POST requests to the evaluator service."""
-        response = requests.post(url=f"{self.config.evaluator_service_url}/{path}", json=data)
+        response = requests.post(url=f"{self.config.evaluator_service_url}{path}", json=data)
         response.raise_for_status()
         return response.json()
 

llama_stack/providers/remote/post_training/nvidia/post_training.py

@@ -408,7 +408,7 @@ class NvidiaPostTrainingAdapter(ModelRegistryHelper):
                     if v is not None
                 }
             else:
-                raise NotImplementedError(f"JASH was here Unsupported algorithm config: {algorithm_config}")
+                raise NotImplementedError(f"Unsupported algorithm config: {algorithm_config}")
 
         # Create the customization job
         response = await self._make_request(