We’re trying our best to keep Prodigy typo-free, but they still creep in sometimes. So I thought I’d create a master thread to make it easier to report small typos and inconsistencies in the documentation, website, commands, logging etc.
Thanks to everyone who’s been taking the time to report minor issues like this so far. Don’t ever feel bad or pedantic about reporting small typos btw – we love attention to detail!
When I try to include a TextClassifier in a custom recipe my interpreter barfs:
import spacy
from prodigy.models.ner import EntityRecognizer
from prodigy.models.textcat import TextClassifier
nlp = spacy.load('en_core_web_lg')
ner_model = EntityRecognizer(nlp, label=['TACO', 'BURRITO']) # works
cat_model = TextClassifier(nlp, label=['TACO']) # fails
The error message is:
cython_src/prodigy/models/textcat.pyx in prodigy.models.textcat.TextClassifier.__init__()
TypeError: __init__() takes at least 3 positional arguments (2 given)
My environment has prodigy==1.1.0 and spacy==2.0.5.
Mostly unrelated (and trivial) - I think in PRODIGY_README.html#models there is a typo in the EntityRecognizer section where prodigy.components.ner.EntityRecognizer should be prodigy.models.ner.EntityRecognizer.
Prodigy 1.4.0. In the recipe_args dictionary in core.py at line 594 there is the following definition:
'whole_text': (
'Make accept/reject refer to whole text whole text (not single span)',
'flag',
'W',
'<value is a self-reference, replaced by this string>',
),
The “a n” for “an” typo for ner.eval is in the help documentation as well as the web documentation.
$ pgy ner.eval --help
usage: prodigy ner.eval [-h] [-a None] [-lo None] [-l None] [-e None] [-W]
[-U]
dataset model [source]
Evaluate a n NER model and build an evaluation set from a stream.
Insanely small. Some of the documentation is missing arguments. For example, the section in the documentation for db-out (PRODIGY_README.html#manage-db-out) is missing the –answer/-a description.
I realized it accepted this argument when I looked at the command line help section:
➜ prodigy db-out -h
usage: prodigy db-out [-h] [-a None] [-F] [-D] set_id [out_dir]
Export annotations from the database. Files will be exported in
Prodigy's JSONL format.
positional arguments:
set_id Dataset ID
out_dir Path to output directory
optional arguments:
-h, --help show this help message and exit
-a None, --answer None
Only export annotations with this answer
-F, --flagged-only Only export flagged annotations
-D, --dry Perform a dry run