In function "resolve_tokenizer", the program will always connect to Huggingface to get autoprocessor, this might slow down the speed and cause problem when there is no well connection to the public internet or Huggingface. So will it be better to download the autoprocessor with the model, and allow the program to load local autoprocessor after the first downloading?
In function "resolve_tokenizer", the program will always connect to Huggingface to get autoprocessor, this might slow down the speed and cause problem when there is no well connection to the public internet or Huggingface. So will it be better to download the autoprocessor with the model, and allow the program to load local autoprocessor after the first downloading?