Update app.py

app.py

@@ -1,6 +1,11 @@
 import gradio as gr
-import json
-import os
+import subprocess
+# Install flash attention, skipping CUDA build if necessary
+subprocess.run(
+    "pip install flash-attn --no-build-isolation",
+    env={"FLASH_ATTENTION_SKIP_CUDA_BUILD": "TRUE"},
+    shell=True,
+)
 from huggingface_hub import snapshot_download 
 snapshot_download(repo_id="jzq11111/mooncast", local_dir='./resources/')