This repo is a modified version of https://github.com/jacobgil/pytorch-pruning.
This is an implementation of the paper [1611.06440 Pruning Convolutional Neural Networks for Resource Efficient Inference].
This demonstrates pruning a VGG16 based classifier that classifies a small dog/cat dataset.
This was able to reduce the CPU runtime by x3 and the model size by x4.
For more details you can read the blog post.
At each pruning step 512 filters are removed from the network.
Tested Environment:
- CUDA 9.1
- Python 3.6 with PyTorch
This repository uses the PyTorch ImageFolder loader, so it assumes that the images are in a different directory for each category.
├── dataset.py
├── finetune.py
├── prune.py
├── test
│ ├── cats
│ └── dogs
└── train
├── cats
└── dogs
The images were taken from here but you should try training this on your own data and see if it works!
Training:
python finetune.py --train
Pruning:
python finetune.py --prune
-
Change the pruning to be done in one pass. Currently each of the 512 filters are pruned sequentually.
for layer_index, filter_index in prune_targets: model = prune_vgg16_conv_layer(model, layer_index, filter_index)This is inefficient since allocating new layers, especially fully connected layers with lots of parameters, is slow.
In principle this can be done in a single pass.
-
Change prune_vgg16_conv_layer to support additional architectures. The most immediate one would be VGG with batch norm.