search

rinongal / textual_inversion

MIT License
2.87k stars 278 forks source link

max_vectors_per_token #139

Closed fido20160817 closed 1 year ago

fido20160817 commented 1 year ago

what does 'max_vectors_per_token ' means? under what kind of situation? any example?

rinongal commented 1 year ago

It was used as part of the experiment in the paper where we progressively add more vectors to the inversion. Probably not relevant to you unless you set progressive_words to True in the config.

fido20160817 commented 1 year ago

Thanks for your response. Another stupid question, what does 'progressive_words' means?

rinongal commented 1 year ago

We describe this experiment in the paper. It is equivalent to starting training with only a single word representing your concept and adding more and more words as training progresses

fido20160817 commented 1 year ago

many thanks!