ZIYU-DEEP/quantum-sparse-training

😈 READ ME!

1. Intro

This repository is organized as follows:

sparse-training-via-information
|== loader
│   └── loader_cifar100.py
│   └── loader_cifar100_noisy.py
│   └── main.py
|== network
│   └── mlp.py
│   └── resnet.py
│   └── main.py
|== optim
│   └── trainer.py
│   └── model.py
|== helper
│   └── pruner.py
│   └── utils.py
|== scripts
│   └── cifar100_resnet.sh
|== run.py
|== experiment.py

2. Requirements

If you are using anaconda:

conda create --name sparse python=3.8
conda activate sparse

To install necessary pakacges, check the list in ./requirements.txt or lazily run the following in the designated environment for the project:

python3 -m pip install -r requirements.txt

3. Example Commands

# Training the network
. scripts/train.sh

# Get the Hessian spectrum
. scripts/spectrum.sh

Be sure check the original file for more detailed instructions on datasets, network, and optimization options, and modify the bash scripts accordingly. I have written help strings for every argument.