Awesome

DPFPS: Dynamic and Progressive Filter Pruning for Compressing Convolutional Neural Networks from Scratch

The PyTorch implementation of DPFPS (AAAI 2021).

Requirements

Ubuntu Version: 18.04.2 LTS;

Python Version: 3.6.9;

Pytorch Version: 1.2.0;

CUDA Version: 10.1.243.

Run

Test

The pruned models can be found in Google Drive.

python pruned_models_test.py --arch $architecture$  --model $pruned models$ --dataset $dataset$

Train

bash ./scripts/run_resnet101_dpss_pr_0.45.sh