feature: Native SpaCy runner for v1.0.0 architecture

[connorbrinton]

Feature request

BentoML previously supported SpaCy as one of its built-in runner types. However, with the introduction of the v1.0.0 release, native support for SpaCy was removed.

It would be great to have native support for a built-in SpaCy runner 🙂

Motivation

We've successfully deployed BentoML for a PyTorch-based image classification application, but also heavily use SpaCy in production. We'd like to move these SpaCy-based services over to BentoML, and having native support for SpaCy in BentoML would make this easier.

As a workaround, we plan to implement a custom runner for SpaCy for our use case. We'll try to reuse code from the previous implementation as much as possible.

Other

Since we'll already be writing a (custom) SpaCy runner, we may be able to contribute our implementation to BentoML if that would be of interest

[aarnphm]

Happy to review if you submit a PR

[SergheiDinu]

Did u manage to have one?