Reducing the size of the Docker image
This is more a question than an issue:
Is it possible to
- install the forked AllenNLP repository without dependencies? (Do we use the functionality or just the data? If functionality, why do we not use the upstream repository?)
- move model data to it's own sub-directory under the root directory and load it from there?
- Additionally, could the location be configured via an environment variable?
Edited by Manuel Prinz