search

NVlabs / DeepInversion

Official PyTorch implementation of Dreaming to Distill: Data-free Knowledge Transfer via DeepInversion (CVPR 2020)
Other
485 stars 77 forks source link

VGG Architectures #13

Closed derekwan1 closed 2 years ago

derekwan1 commented 3 years ago

Hi there, What are the dimensions of the fully connected layers for the VGG's used in this paper? Thanks!

hongxuyin commented 3 years ago

Hi, FC dims at 512-10 for CIFAR-10, and 4096-1000 for ImageNet.