feat: Support "stop" parameter in remote:vLLM

Signed-off-by: Yuan Tang <terrytangyuan@gmail.com>

llama_stack/models/llama/datatypes.py

@@ -201,6 +201,7 @@ class SamplingParams(BaseModel):
 
     max_tokens: Optional[int] = 0
     repetition_penalty: Optional[float] = 1.0
+    stop: Optional[List[str]] = None
 
 
 class CheckpointQuantizationFormat(Enum):

llama_stack/providers/utils/inference/openai_compat.py

@@ -147,6 +147,9 @@ def get_sampling_options(params: SamplingParams) -> dict:
         if params.repetition_penalty is not None and params.repetition_penalty != 1.0:
             options["repeat_penalty"] = params.repetition_penalty
 
+        if params.stop is not None:
+            options["stop"] = params.stop
+
     return options