Update app.py

app.py

@@ -939,7 +939,7 @@ if __name__ == "__main__":
         attn_implementation='flash_attention_2',  # OR 'sdpa' for Ascend NPUs
         # torch_dtype=args.torch_dtype,
         # attn_implementation='sdpa',  # OR 'sdpa' for Ascend NPUs
-        token=os.environ["HF_TOKEN"],   
+        use_auth_token=True,   
         low_cpu_mem_usage=True,
         trust_remote_code=True).eval().cuda()
     processor = AutoProcessor.from_pretrained(args.model_name, trust_remote_code=True)
@@ -948,13 +948,13 @@ if __name__ == "__main__":
     dualvitok = AutoModel.from_pretrained(args.tokenizer_path,
                                           torch_dtype=torch.float32,
                                           trust_remote_code=True,
-                                          token=os.environ["HF_TOKEN"],   
+                                          use_auth_token=True,   
                                           ).eval().cuda()
     processor.set_vision_tokenizer(dualvitok)
 
     # (Optional): set the sdxl diffusion decoder. It will enable upsample 2x image resolution.
     processor.load_diffusion_vision_detokenizer(args.diffusion_decoder_path,
-                                                token=os.environ["HF_TOKEN"])
+                                                use_auth_token=True)
 
     # Assign device to model for later use
     streamer = TextIteratorStreamer(processor.tokenizer, skip_prompt=True, skip_special_tokens=True, timeout=15)