CSS 490a Acceleration Project

Aiming to accelerate neural network inference time on embedded systems via model compression, specifically pruning and quantization.

Setup

# Set up Python environment
virtualenv --python $(which python3) venv
source venv/bin/activate
pip install -r requirements.txt

# Download dataset
sh download_dataset.sh

Usage

# classify image with unmodified VGG-16
python run_classify.py

# prune model
python prune_model.py