# ResNeXt152_32x4d python train.py \ --model=ResNeXt152_32x4d \ --batch_size=256 \ --lr_strategy=piecewise_decay \ --lr=0.1 \ --num_epochs=120 \ --model_save_dir=output/ \ --l2_decay=1e-4
