Tips to reduce memory requirements for fastai model during inference inside docker

fastai
2

How are you launching the docker image? Generally you should be able to raise the shard memory space of the docker image when launching it.

E.g. you can pass:

--shm-size=24G