Possible to disable pipeline component by default?

[kinghuang]

It's possible to disable a pipeline component on a loaded model using Language.disable_pipe and similar methods. But, is it possible to configure a pipeline component to be disabled by default on load? I would like to include a pipeline component that is disabled by default, and has to be explicitly enabled.

The effect would be something like this, where a model has a pipeline component named some_pipe that has to be explicitly enabled.

nlp = spacy.load("some_model")
print(nlp.pipe_names)

['tok2vec', 'tagger', 'parser', 'attribute_ruler', 'lemmatizer']

nlp.enable_pipe("some_pipe")
print(nlp.pipe_names)

['tok2vec', 'tagger', 'parser', 'attribute_ruler', 'lemmatizer', 'some_pipe']

[adrianeboyd]

Sure, just disable the component before saving the pipeline, and it should be disabled by default on load.

The trained pipeline core packages all include a disabled senter, if it helps to see a concrete example of what the config looks like.

[kinghuang]

Whoa, I've never noticed the disabled senter component in the core packages before! 🤯

Thanks for pointing out the core configs. It made me realize there's a disabled config that takes in a list of disabled component names.

pipeline = ["tok2vec","tagger","parser","attribute_ruler","lemmatizer","ner","llm_ner"]
disabled = ["llm_ner"]