textcat.batch-train documentation fix?

, ,
1

Hi,

The recipe for textcat.batch-train shows usage with a --label parameter:

prodigy textcat.batch-train gh_issues /tmp/model --n-iter 10 --eval-split 0.2 
--dropout 0.2 --label DOCUMENTATION

However, this usage returns an error for me:

python -m prodigy textcat.batch-train newsprocessor_category en_core_web_md --label DOCUMENTATION
usage: prodigy textcat.batch-train [-h] [-o None] [-la en] [-f 1] [-d 0.2]
                                   [-n 10] [-b 10] [-e None] [-es None] [-L]
                                   [-S]
                                   dataset [input_model]
prodigy textcat.batch-train: error: unrecognized arguments: --label DOCUMENTATION

Looks like the --label argument is not supported. Similar issue seems to have been reported earlier for workflow documentation:

2

Ah thanks, will fix! And yes, the --label argument used to be necessary, but the recipe now reads the labels from the data automatically.

3

Thankyou! I’ve met the same problem. But I think ‘–label’ for a textcat mission is pretty necessary. For example I have annotated some samples for 7 categories but now I just want to train and save a model for only ONE of those categories , without --label which allows me to choose the cate I want, things can be inconvenient.