Update app.py

app.py

@@ -1,15 +1,20 @@
 import gradio as gr
 import os 
 from pathlib import Path
+
+os.environ["CMAKE_ARGS"] = "-DLLAMA_CUBLAS=on"
+os.system('CMAKE_ARGS="-DLLAMA_CUBLAS=on" pip install llama-cpp-python[server]')
+
 import argparse
-model_file = "yi-chat-6b.Q3_K_M.gguf"
+model_file = "yi-chat-6b.Q4_K_M.gguf"
 if not os.path.isfile(model_file):
-    os.system("wget -c https://huggingface.co/XeIaso/yi-chat-6B-GGUF/blob/main/yi-chat-6b.Q3_K_M.gguf")
+    os.system("wget -c https://huggingface.co/XeIaso/yi-chat-6B-GGUF/resolve/main/yi-chat-6b.Q4_K_M.gguf")
 
 DEFAULT_MODEL_PATH = model_file
 
 from llama_cpp import Llama
 llm = Llama(model_path=model_file, model_type="mistral")
+llm._token_eos = 7
 
 
 def predict(input, chatbot, max_length, top_p, temperature, history):
@@ -17,7 +22,7 @@ def predict(input, chatbot, max_length, top_p, temperature, history):
     response = ""
     history.append(input)
 
-    for output in llm(input, stream=True, temperature=temperature, top_p=top_p, max_tokens=max_length, ):
+    for output in llm(input, stream=True, temperature=temperature, top_p=top_p, max_tokens=max_length, stop=["<|im_end|>"]):
         piece = output['choices'][0]['text']
         response += piece
         chatbot[-1] = (chatbot[-1][0], response)
@@ -37,7 +42,7 @@ def reset_state():
 
 
 with gr.Blocks() as demo:
-    gr.HTML("""<h1 align="center">Yi-6B-GGUF by llama-cpp-python</h1>""")
+    gr.HTML("""<h1 align="center">Yi-6B-Chat by llama-cpp-python</h1>""")
 
     chatbot = gr.Chatbot()
     with gr.Row():