tokenize not add bos

app.py

@@ -7,22 +7,11 @@ if not os.path.isfile("Yi-6B.q4_k_m.gguf"):
     os.system("wget -c https://huggingface.co/SamPurkis/Yi-6B-GGUF/resolve/main/Yi-6B.q4_k_m.gguf")
 
 DEFAULT_MODEL_PATH = model_file
-parser = argparse.ArgumentParser()
-parser.add_argument("-m", "--model", default=DEFAULT_MODEL_PATH, type=Path, help="model path")
-parser.add_argument("--mode", default="chat", type=str, choices=["chat", "generate"], help="inference mode")
-parser.add_argument("-l", "--max_length", default=512, type=int, help="max total length including prompt and output")
-parser.add_argument("-c", "--max_context_length", default=512, type=int, help="max context length")
-parser.add_argument("--top_k", default=0, type=int, help="top-k sampling")
-parser.add_argument("--top_p", default=0.7, type=float, help="top-p sampling")
-parser.add_argument("--temp", default=0.95, type=float, help="temperature")
-parser.add_argument("--repeat_penalty", default=1.1, type=float, help="penalize repeat sequence of tokens")
-parser.add_argument("-t", "--threads", default=0, type=int, help="number of threads for inference")
-parser.add_argument("--plain", action="store_true", help="display in plain text without markdown support")
-args = parser.parse_args()
 
 from llama_cpp import Llama
 llm = Llama(model_path=model_file)
-
+old_tokenize = llm._model.tokenize
+llm._model.tokenize = lambda text, add_bos, spec: old_tokenize(text, False, spec)
 
 
 def predict(input, chatbot, max_length, top_p, temperature, history):
@@ -50,7 +39,7 @@ def reset_state():
 
 
 with gr.Blocks() as demo:
-    gr.HTML("""<h1 align="center">Yi-6B-GGUF by llama.cpp</h1>""")
+    gr.HTML("""<h1 align="center">Yi-6B-GGUF by llama-cpp-python</h1>""")
 
     chatbot = gr.Chatbot()
     with gr.Row():
@@ -58,9 +47,9 @@ with gr.Blocks() as demo:
             user_input = gr.Textbox(show_label=False, placeholder="Input...", lines=8)
             submitBtn = gr.Button("Submit", variant="primary")
         with gr.Column(scale=1):
-            max_length = gr.Slider(0, 32048, value=args.max_length, step=1.0, label="Maximum Length", interactive=True)
-            top_p = gr.Slider(0, 1, value=args.top_p, step=0.01, label="Top P", interactive=True)
-            temperature = gr.Slider(0, 1, value=args.temp, step=0.01, label="Temperature", interactive=True)
+            max_length = gr.Slider(0, 32048, value=2048, step=1.0, label="Maximum Length", interactive=True)
+            top_p = gr.Slider(0, 1, value=0.7, step=0.01, label="Top P", interactive=True)
+            temperature = gr.Slider(0, 1, value=0.95, step=0.01, label="Temperature", interactive=True)
             emptyBtn = gr.Button("Clear History")
 
     history = gr.State([])