How to run single component on Doc?

[wildfluss]

I save my Doc to file like this

nlp = en_core_web_lg.load()
doc = nlp(text)
doc_bytes = doc.to_bytes()

and then I want to continue where I left off but I don't want to run the whole pipeline again, but only run my custom component on the deserialized Doc object from disk, where should I start to do it spaCy way eg without importing my custom component implementation function but calling only spaCy APIs if possible ?

[thomashacker]

Hello,

You can use nlp.get_pipe (docs) do get individual components from a spaCy pipeline and use them to process Doc objects.

nlp = spacy.load("your_model")

# doc_bytes = The way you're loading in your bytes_file
doc = Doc(nlp.vocab).from_bytes(doc_bytes)

your_component = nlp.get_pipe("your_component")
result = your_component(doc)

You can also disable or exclude components from a pipeline when using spacy.load() (docs).

spacy.load("your_model", disable=["component_to_disable"]) # Components that will be loaded but do not process any Docs
spacy.load("your_model", exclude=["component_to_exclude"]) # Components that won't be loaded into the model
spacy.load("your_model", enable=["your_component"]) # Disables every component except the components you specified