Update app_v3.py

app_v3.py

@@ -20,6 +20,8 @@ st.write(f'{os.listdir(quantized_model_dir)}')
 model_name_or_path = quantized_model_dir
 model_basename = "Jackson2-4bit-128g-GPTQ"
 
+os.environ['CUDA_VISIBLE_DEVICES']='0'
+
 use_triton = False
 
 tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True, legacy=False)