Model Architecture

[rkoystart]

Hi is there a way to view the layers present in a spacy model (en_core_web_sm). So my question is, In pytorch the following statement
print(model)
will print the architecture (layers within the model), which gives us an easy undertanding on how the input is being transformed or operated.

[ljvmiranda921]

Hi @rkoystart , you can use something like nlp.pipe_names to see the components that underlie a particular spaCy model. You can couple this with nlp.get_pipe_config to see how that particular component was configured. For example:

import spacy

nlp = spacy.load("en_core_web_lg")
for pipe in nlp.pipe_names:
    print(nlp.get_pipe_config(pipe))

You can then cross-reference them with the architectures in the spaCy documentation. Note that a spaCy pipeline will not always contain only neural models, so this method is not exactly a one-to-one comparison with pytorch (that gives you the layers).