llama adapter doesn't support use_cache yet

llama_lora/ui/inference_ui.py

@@ -111,6 +111,7 @@ def do_inference(
             top_k=top_k,
             repetition_penalty=repetition_penalty,
             num_beams=num_beams,
+            use_cache=False,
         )
 
         generate_params = {