I am currently trying to speed up my application by removing extra spaCy component when they are not needed and enabling them at later point of time. I have come-up with this code.
import spacy
nlp = spacy.load("en_core_web_lg", disable=('ner', 'textcat'))
nlp.pipe_names
It given the following output
['tagger', 'parser']
I have to perform a task, below is the code snippet
text = """Extracts the selected layers in the specified area of interest.... """
doc = nlp(text)
def get_pos(remove_parser=True):
if remove_parser:
nlp.remove_pipe("parser")
for kw in keywords:
doc = nlp(kw[0])
tag_list = [(token.text, token.tag_) for token in doc]
if remove_parser:
nlp.add_pipe(nlp.create_pipe('parser'))
return tag_list
result = get_pos(remove_parser=True)
nlp.pipe_names
So I call the get_pos function with remove_parser=True. It removes the parser component, run nlp(kw[0]) for every item in keywords list. After the loop ends I add back the parser component, which can be verified by the output of the nlp.pipe_names command. I get the below output
['tagger', 'parser']
But then if I call nlp("Hello World") after the get_pos function call. It gives this error -
ValueError Traceback (most recent call last)
<ipython-input-29-320b76b1fe36> in <module>
----> 1 nlp("Hello World")
~\.conda\envs\keyword-extraction\lib\site-packages\spacy\language.py in __call__(self, text, disable, component_cfg)
433 if not hasattr(proc, "__call__"):
434 raise ValueError(Errors.E003.format(component=type(proc), name=name))
--> 435 doc = proc(doc, **component_cfg.get(name, {}))
436 if doc is None:
437 raise ValueError(Errors.E005.format(name=name))
nn_parser.pyx in spacy.syntax.nn_parser.Parser.__call__()
nn_parser.pyx in spacy.syntax.nn_parser.Parser.predict()
nn_parser.pyx in spacy.syntax.nn_parser.Parser.require_model()
ValueError: [E109] Model for component 'parser' not initialized. Did you forget to load a model, or forget to call begin_training()?
You are trying to add a blank/untrained parser back to the pipeline rather the one that was provided with it. Instead, try disable_pipes(), which makes it easier to save the component and add it back later:
disabled = nlp.disable_pipes(["parser"])
# do stuff
disabled.restore()