run_eval_multi_xl.sh

#CUDA_VISIBLE_DEVICES=4,5,6,7 \
#python main_multi_gpu.py \
GLOG_v=0 python3 -m paddle.distributed.launch --gpus="0,1,2,3,4,5,6,7" main_multi_gpu.py \
-cfg='./configs/replknet_xl.yaml' \
-dataset='imagenet2012' \
-batch_size=2 \
-data_path='/dataset/imagenet' \
-pretrained='./replknet_xl_73m_to_1k_320.pdparams' \
-eval \
-amp