Closed aCampello closed 1 year ago
@nsorros
Totally agree. I wonder to which extent two repos are needed. One for experimentation and one for production with different "requirements", one with more libraries and less optimised, the other less heavy.
Otherwise it could happen in this repo as well but the code would need to be decoupled and we would need maybe a light setup.py and a more heavy requirements.txt to be used for experimentation
We are addressing that in #246
At the moment, to run any model (e.g. xlinear) in production, we need to install the whole package. This is probably not ideal, and blows up the size of the docker container. It would be good to have a minimal code just for a production model.