Does anyone know how the provided VGG16_bn model was trained? I want to train the VGG16 with batch normalization from scratch for some use.
I guess the code is
Thanks for you reply @chenyuntc . It is my mistake. Actually I want to know the hyper-parameters such as the batch size, learning-rate, epochs, etc. The default parameters do not work.
Sorry but I can’t find the origin code to reproduce vgg16-bn result.
I could only find a related issue here(@colesbury is responsible for vgg in torchvision )