Skip to content

Latest commit

 

History

History
20 lines (12 loc) · 466 Bytes

File metadata and controls

20 lines (12 loc) · 466 Bytes

Task-Oriented Feature Distillation

This is the implementation of Task-Oriented Feature Distillation, NeurIPS2020.

Experiments on CIFAR100

Step1. Install the required packages.

pip install torch torchvision

Step2. Train a student model.

python distill.py --model=resnet18

Note that you can choose resnet, senet and preactresnet models as students. Each model has five kinds of depth - 18, 34, 50, 101, and 152.