change to dev, fix issues with test

llama_stack/providers/remote/inference/centml/centml.py

@@ -256,32 +256,6 @@ class CentMLInferenceAdapter(
         logcat.debug("inference", f"params to centml: {params}")
         return params
 
-    def _build_options(
-        self,
-        sampling_params: Optional[SamplingParams],
-        fmt: Optional[ResponseFormat],
-    ) -> dict:
-        """
-        Build temperature, max_tokens, top_p, etc., plus any response format data.
-        """
-        options = get_sampling_options(sampling_params)
-        options.setdefault("max_tokens", 512)
-
-        if fmt:
-            if fmt.type == ResponseFormatType.json_schema.value:
-                options["response_format"] = {
-                    "type": "json_object",
-                    "schema": fmt.json_schema,
-                }
-            elif fmt.type == ResponseFormatType.grammar.value:
-                raise NotImplementedError(
-                    "Grammar response format not supported yet"
-                )
-            else:
-                raise ValueError(f"Unknown response format {fmt.type}")
-
-        return options
-
     def _build_options(
         self,
         sampling_params: Optional[SamplingParams],
@@ -302,7 +276,7 @@ class CentMLInferenceAdapter(
                 raise ValueError(f"Unknown response format {fmt.type}")
 
         if logprobs and logprobs.top_k:
-            options["logprobs"] = 1
+            options["logprobs"] = logprobs.top_k
 
         return options
 

llama_stack/templates/centml/run.yaml

@@ -16,7 +16,7 @@ providers:
   - provider_id: centml
     provider_type: remote::centml
     config:
-      url: https://api.centml.com/openai/v1
+      url: https://api.centml.org/openai/v1
       api_key: "${env.CENTML_API_KEY}"
   - provider_id: sentence-transformers
     provider_type: inline::sentence-transformers
@@ -104,6 +104,12 @@ metadata_store:
   db_path: ${env.SQLITE_STORE_DIR:~/.llama/distributions/centml}/registry.db
 
 models:
+- metadata: {}
+  model_id: meta-llama/Llama-3.1-3B-Instruct
+  provider_id: centml
+  provider_model_id: meta-llama/Llama-3.2-3B-Instruct
+  model_type: llm
+
 - metadata: {}
   model_id: meta-llama/Llama-3.3-70B-Instruct
   provider_id: centml