Update app.py

app.py

@@ -5,6 +5,12 @@ import gradio as gr
 import time
 import os
 
+subprocess.run(
+    "pip install flash-attn --no-build-isolation",
+    env={"FLASH_ATTENTION_SKIP_CUDA_BUILD": "TRUE"},
+    shell=True,
+)
+
 MB = 1024*1024
 def eat_memory():
     mem = []