ELAN：将超分网络SwinIR高效化，

https://github.com/xindongzhang/ELAN

pip install pytorch-msssim -i https://pypi.tuna.tsinghua.edu.cn/simple
pip install pyyaml -i https://pypi.tuna.tsinghua.edu.cn/simple
pip install tqdm -i https://pypi.tuna.tsinghua.edu.cn/simple

cd ELAN
python train.py --config ./configs/elan_light_x4.yml

data_path: '/home/xindongzhang/SR_datasets'

data_path: './DIV2K_train_HR'
eval_sets: ['Set5', 'Set14', 'B100', 'Urban100', 'Manga109']

gpu_ids: [4, 5, 6, 7]

batch_size: 64

改为

data_path: './train_data'
eval_sets: ['Set5']
gpu_ids: [0]

batch_size: 8

设置数据集路径：

创建文件夹：

\train_data\div2k_cache\div2k_hr\rgb\DIV2K_train_HR

train_data\div2k_cache\div2k_lr_x4\rgb\DIV2K_train_LR_bicubic\X4

train_data\div2k_cache\div2k_lr_x2\rgb\DIV2K_train_LR_bicubic\X2

train_data\\benchmark/Set5/HR

继续训练：

python train.py --config ./configs/elan_light_x2.yml --resume ./experiments/elan-fp32-x2-2022-0718-1936