Update app.py

app.py

@@ -1,6 +1,9 @@
 import time
 from threading import Thread
 
+import subprocess
+subprocess.run(["pip", "install", "."])
+
 import gradio as gr
 from io import BytesIO
 import requests
@@ -11,7 +14,6 @@ from llava.mm_utils import tokenizer_image_token
 from transformers.generation.streamers import TextIteratorStreamer
 import spaces
 
-import subprocess
 
 if torch.cuda.is_available():
     device = "cuda"
@@ -20,8 +22,6 @@ else:
     device = "cpu"
     print("Using CPU")
 
-subprocess.run(["pip", "install", "."])
-
 device = "cuda:0"
 tokenizer, model, image_processor, context_len = load_pretrained_model(model_path="./checkpoints/llava-phi3-3.8b-lora", model_name="llava-phi3-3.8b-lora", model_base="microsoft/Phi-3-mini-128k-instruct", load_8bit=False, load_4bit=False, device=device)
 model.to(device)