Torchvision: finding details on how pretrained models were pretrained

I am looking for details on exactly how the torchvision models are pretrained so I can train these models on the CIFAR data sets, rather than the current ImageNet.

Sadly the training scripts for the pretrained models do not seem to be included with the torchvision source and many details are missing from the documentation.

However, I did find this script. Is it the script used for pretraining? And if not, where I can find this info?


1 Like

Bump. Iā€™m interested in the resnets specifically