Update app.py

app.py

@@ -25,7 +25,7 @@ device = args.device
 assert device in ['cuda', 'mps']
 
 # Load model
-model = AutoModel.from_pretrained('openbmb/MiniCPM-Llama3-V-2_5-int4', trust_remote_code=True)
+model = AutoModel.from_pretrained('openbmb/MiniCPM-Llama3-V-2_5-int4', trust_remote_code=True, load_in_4bit=True)
 tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
 model.eval()