Evaluation took so much disk

[uluumbch]

I am try to train SegFormer model and use evaluate module to eval the IoU, but when there is so many data and epoch, the metrics looks like take so much space.

is there any method so I can free space while training still going?