Ram (not cuda) was increasing while training until it was forced to kill, not sure how to fix it?

[littlewhite152]

When I use the voc dataset it works fine, but when I switch back to the coco dataset it happens.

[Isalia20]

Try to use less num_workers or num_workers=0 in config file. Include this in training config file: DATA_LOADER: NUM_WORKERS: 0