Support captions list in preprocess files

[A2va]

With this change, it's possible to preprocess files with manual captioning

[A2va]

Also I have a question, in cli_lora_pti.py we can add more than one concept (person/style/object) but how to train two concepts with the same data set ? Do I need to caption files with the concept token ?

[cloneofsimo]

Thanks! You can optionally provide captions and use token_map to train two concepts