I want to pack my model to docker image.
Cause i’m beginner in domain of docker, i just know one solution - make a REST API server with pytorch model and make it as docker image. Then, when use this docker server, send request to train or test model.
Is there other options??
That workflow is similar to how
torchserve users have deployed a model in the past, check it out and lemme know if you have any questions https://github.com/pytorch/serve/tree/master/docker but this should make it easy to take a trained model, serve it from within a container and expose an HTTP endpoint that users can query against
If you’re looking at packaging a model in Docker file for training
Then in your
Dockerfile you need to
ADD directory_with_data_and_code and then
RUN python train.py would be an easy start