Pytorch training deep learning network settings CUDA specified GPU visible

cuda designated GPU so

• The original training command is ：python train.py --config config/pascal/pascal_split0_resnet50.yaml , But it will use GPU All graphics cards on
• Want to use the specified graphics card to train , Can be in their own python preorder ：CUDA_VISIBLE_DEVICES=0,1, here 0,1 It's your own graphics card number
• CUDA_VISIBLE_DEVICES=0,1 python train.py You can use two cards 0,1 Normal training
  

！！！ Note that there CUDA_VISIBLE_DEVICES=0,1, here CUDA_VISIBLE_DEVICES、 Equal sign and 0,1 There must be no space between them , Otherwise it will warn ：WARNING:root:could not open file '/etc/apt/sources.list.d, And report an error