Merge pull request #761 from kvcache-ai/fix-server-bug

fix ollama api temperature bug

ktransformers/server/backend/interfaces/ktransformers.py

@@ -227,7 +227,7 @@ class KTransformersInterface(TransformersInterface):
         device = self.device_map.get("blk.0.self_attn", {}).get("generate_device", "cuda:0")
         return torch.tensor([self.seq_length - 1], device=device)
     
-    async def inference(self, local_messages, thread_id: str, temperature: Optional[float], top_p: Optional[float]):
+    async def inference(self, local_messages, thread_id: str, temperature: Optional[float] = None, top_p: Optional[float] = None):
         async with self._infer_lock:
             async for v in super().inference(local_messages, thread_id, temperature, top_p):
                 yield v