解决GPU--CPU转换以及加载多GPU模型后使用pytorch的DataParallel()时出现的错误
1、解决runtimeError:AttemptingtodeserializeobjectonaCUDAdevicebuttorch.cuda.is_available()isFalse.IfyouarerunningonaCPU-onlymachine,pleaseusetorch.loadwithmap_location=‘cpu’tomapyourstoragestotheCPU.参考链接