Jina AI org
edited Sep 17

What I'm proposing to change and why:

  • xlm-r languages from 89 to 100, both paper and model page say they trained it on 100 languages
  • in the mean pooling example, model should receive adapter mask rather than task. task goes to encode() function.
  • truncate dim can not be any integer, it should be one of the following (32, 64, 128, 256, 512, 768, 1024) which is specified in the top section of readme. also, it is not really related with sequence lengths.
jupyterjazz changed pull request status to open
Jina AI org

LGTM

jupyterjazz changed pull request status to merged

Sign up or log in to comment