Preparing

Put all datasets under ./data.

Pretrained Backbone:

ResNet-50 | ResNet-101

├── ./pretrained
    ├── resnet50.pth
    └── resnet101.pth

Training Scripts

# use torch.distributed.launch
sh tools/dist_train.sh <config> <port> <save_path> <num_gpu>

# use slurm
GPUS=<num_gpu> GPUS_PER_NODE=<num_gpu> CPUS_PER_TASK=<num_cpu> sh tools/slurm_train.sh <partition> <config> <port> <save_path>