Update app.py

app.py

@@ -23,10 +23,9 @@ quant_config = BitsAndBytesConfig(
 tokenizer = LlamaTokenizer.from_pretrained(model_name)
 model = LlamaForCausalLM.from_pretrained(
     model_name,
-    quantization_config=quant_config,
-    device_map="auto",
+    quantization_config=quant_config
 )
-model = PeftModel.from_pretrained(model, lora_name)
+model = PeftModel.from_pretrained(model, lora_name).to("cuda")
 
 
 # will be used with normal template