Update app.py

app.py

@@ -55,6 +55,7 @@ model = AutoModelForCausalLM.from_pretrained(
     model_name,
    # quantization_config=quantization_config_4bit, # Comment out if not using quantization
     device_map="auto",
+    offload_folder='./',
 ).to(torch.float16) #.to(torch.device("cuda:0"), torch.bfloat16)
 
 print(f"Loading tokenizer: {model_name}")