Is there any way to cache the data when training with 'ddp' ?

[seungtaek94]

Hi,

I try to cache the data for speed up training.

when I using only pytorch, I cache the data like below.

Its use share the data across the process using shared_dict.

https://github.com/ptrblck/pytorch_misc/blob/31ac50c415f16cf7fec277dbdba72b9fb4d732d3/shared_dict.py

Is there any way to cache the data in pytorch-lighting similar to above?