Extend Pipeline using Config File

[maximilianreimer]

I would like to extend an existing trained pipeline and be able to use the command-line interface to run things like an evaluation on it.

Is there a way of doing this?

So the code looks currently like this:

nlp = spacy.load("en_core_web_trf")
nlp.add_pipe("sentencizer")

I know that you can evaluate the en_core_web_trf pipeline using python -m spacy evaluate en_core_web_trf data/path. What is the best way to run this for the extended pipeline?

I was hoping that I could create a config file that looks somewhat like this:

#mypipeline.cfg
[nlp]
lang = "en"
pipeline = ["en_core_web_trf","sentencizer"]

and run python -m spacy evaluate mypipeline data/path.

Thanks

[polm]

If you create your pipeline in code like that and save it to disk you can specify it from the command line just like en_core_web_trf. Just pass the path to the model you saved instead of en_core_web_trf.