Training on binary annotations throws error

,
1

Problem - Trying to train a model that was annotated using ner.teach command throws up error.

Train command - prodigy ner.batch-train ner_person_binary en_core_web_lg --output OUTPUT_PATH --label PERSON --eval-split 0.2 --n-iter 10 --batch-size 8

ERROR -
self.extend(nlp.remove_pipe(name) for name in names)
File "/opt/anaconda3/lib/python3.8/site-packages/spacy/language.py", line 415, in remove_pipe
raise ValueError(Errors.E001.format(name=name, opts=self.pipe_names))
ValueError: [E001] No component 'tagger' found in pipeline. Available names: ['ner']

prodigy version - '1.10.7'
spacy version - 2.3.0

Let me know if I am missing something here.

2

@ines Hi Ines :slight_smile: , could you help me out here?

3

Could you try upgrading to v1.10.8? I think we fixed an issue in the latest version that could be related to this.

4

Just updated it and still having some issues.

raise ValueError(Errors.E008.format(names=unexpected))
ValueError: [E008] Some current components would be lost when restoring previous pipeline state. If 
you added components after calling `nlp.disable_pipes()`, you should remove them explicitly with 
`nlp.remove_pipe()` before the pipeline is restored. Names of the new components: ['sentencizer']
5

Just released v1.11, which removes the need for the --binary flag and separate annotation models during training and now also supports training from a mix of binary and manual annotations: https://prodi.gy/docs/recipes#train

1 Like