# 直接加载模型
model.load_state_dict(torch.load('./data/my_model.pkl'))
#GPU训练的模型加载到CPU上:
model.load_state_dict(torch.load('./data/my_model.pkl', map_location=lambda storage, loc: storage))
#加载到GPU1上:
model.load_state_dict(torch.load('./data/my_model.pkl', map_location=lambda storage, loc: storage.cuda(1)))
#从GPU1 移动到 GPU0:
model.load_state_dict(torch.load('./data/my_model.pkl', map_location={'cuda:1':'cuda:0'}))