try again

app.py

@@ -8,7 +8,7 @@ import time
 import gradio as gr
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
-hf_token = os.getenv('HF_TOKEN')
+#hf_token = os.getenv('HF_TOKEN')
 
 device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')
 
@@ -26,9 +26,9 @@ PAYLOAD_BITS = 2
 device = torch.device('cuda:0') if torch.cuda.is_available() else torch.device('cpu')
 
 
-model = AutoModelForCausalLM.from_pretrained(args.model, use_auth_token=hf_token, torch_dtype=torch.float16,
+model = AutoModelForCausalLM.from_pretrained(args.model, torch_dtype=torch.float16,
         device_map='auto').to(device)
-tokenizer = AutoTokenizer.from_pretrained(args.model, use_auth_token=hf_token)
+tokenizer = AutoTokenizer.from_pretrained(args.model)
 tokenizer.pad_token = tokenizer.eos_token
 
 DEFAULT_SYSTEM_PROMPT = """\