Integrate koila to avoid CUDA OOM errors

[manisnesan]

@muellerzr Adding this issue based on fastai-dev discussion . I would like to use this issue to learn about callbacks. I liked your idea about having a MemoryMangerCallback callback for this purpose.

https://twitter.com/jeanmarcalkazzi/status/1465418318068781062?s=20 https://github.com/rentruewang/koila

self-assign

[muellerzr]

Hey @manisnesan, how are we doing with this? :)

[manisnesan]

@muellerzr - Spent some time studying about callbacks. I will be making progress on this (mostly this week).

[manisnesan]

One potential issue with koila integration is it requires min python version >=3.8 https://github.com/rentruewang/koila/blob/v0.1.1/setup.py#L28