Closed isabelizimm closed 1 year ago
Currently looking at the spaCy API docs, if using the nlp model itself, each model is from Thinc. The pipeline components are NOT Thinc models, but are using a predict method, much like the support of the other modeling frameworks in vetiver. I'm wondering if it is more useful for people to deploy the Thinc model itself, or the larger pipeline objects created with spaCy. There's also another option where we support Thinc models, which can wrap models like PyTorch or TensorFlow, and then support spaCy pipelines separately.
for NLP, it would be nice to add support for spacy models