Getting the transformers model pre-loaded into the Docker Container

[vvkishere]

I have tried using the saved Bento model in the service.py file. But when ever I build the docker image and run the container, the transformer model used in the pipeline gets downloaded separately.

I can't find a way to run the transformer model locally from the container instead of downloading the file each time from the internet when the container is run.

Any suggestions?

[cirezd]

You could build the docker container and then mount a volume to it that contains your model. I have not tested this with bentoML, but I used this before in a fastapi app and custom Dockerfile.