Update app.py

app.py

@@ -1,6 +1,10 @@
 import gradio as gr
 import os 
 from pathlib import Path
+
+os.environment["CMAKE_ARGS"] = "-DLLAMA_CUBLAS=on"
+os.system('CMAKE_ARGS="-DLLAMA_CUBLAS=on" pip install llama-cpp-python[server]')
+
 import argparse
 model_file = "yi-6b.Q4_0.gguf"
 if not os.path.isfile(model_file):