How to run on multi-node?

[machengcheng2016]

Greetings! @ycliu93
Now I need to run train_net.py on two machines (8 gpus per machine), but I fail to find any useful information in the official repo of detectron2. Could you tell me what is the command, please? Besides, how many times should I run the command as follows:
python3 train_net.py --num-machines 2 --num-gpus 16 --config configs/coco_supervision/faster_rcnn_R_50_FPN_sup1_run1.yaml
And how to set dist_url?
Please help me! Thank you very much.