How to run on a local machine?

[baiyancheng20]

If I have a local dgx2 machine with 4 or 8 gpus, how do I modify the code to run on the local dgx2 machine?

[yaroslavvb]

Check this line from distributed launcher
https://github.com/diux-dev/imagenet18/blob/64944208c9988a5d5b6ebbe3d635f138bf10d1e4/train.py#L188

For instance for 16 machine run it calls the following:

python -m torch.distributed.launch --nproc_per_node=8 --nnodes=16 --node_rank=1 --master_addr=172.31.66.240 --master_port=6006 training/train_imagenet_nv.py /home/ubuntu/data/imagenet --fp16 --logdir /ncluster/runs/imagenet-16.02 --distributed --init-bn0 --no-bn-wd --phases [{‘ep’: 0, ‘sz’: 128, ‘bs’: 64, ‘trndir’: ‘-sz/160’}, {‘ep’: (0, 6), ‘lr’: (1.92, 3.84)}, {‘ep’: 6, ‘bs’: 128, ‘keep_dl’: True}, {‘ep’: 6, ‘lr’: 3.84}, {‘ep’: (11, 14), ‘lr’: (3.84, 1.92)}, {‘ep’: 14, ‘sz’: 224, ‘bs’: 64, ‘trndir’: ‘-sz/352’, ‘min_scale’: 0.087, ‘lr’: 1.92}, {‘ep’: 17, ‘bs’: 224, ‘keep_dl’: True}, {‘ep’: (17, 23), ‘lr’: (1.92, 0.672)}, {‘ep’: (23, 29), ‘lr’: (0.672, 0.0672)}, {‘ep’: 29, ‘sz’: 288, ‘bs’: 128, ‘min_scale’: 0.5, ‘rect_val’: True}, {‘ep’: (29, 37), ‘lr’: (0.0192, 0.0019199999999999998)}, {‘ep’: (37, 40), ‘lr’: 0.0019199999999999998}]

So to run on a single machine, change nnodes=1 and nproc_per_node=8