CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \

python -m torch.distributed.launch --master_addr 127.5.0.9 --master_port 2601 --nproc_per_node=8 \

train_baseline.py \

--model deeplabv3 \

--backbone resnet50_original \

--dataset coco_stuff_164k \

--data [your dataset path]/coco_stuff_164k/ \

--batch-size 16 \

--workers 8 \

--crop-size 512 512 \

--lr 0.02 \

--max-iterations 40000 \

--save-dir [your directory path to store checkpoint files] \

--log-dir [your directory path to store log files] \

--val-per-iters 40000 \

--pretrained-base [your pretrained-backbone path]/resnet50_mixskd.pth

CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \

python -m torch.distributed.launch --master_addr 127.5.0.9 --master_port 2601 --nproc_per_node=8 \

eval.py \

--model deeplabv3 \

--backbone resnet50_original \

--dataset coco_stuff_164k \

--data [your dataset path]/coco_stuff_164k/ \

--save-dir [your directory path to store checkpoint files] \

--pretrained [your pretrained model path]