[NeurIPS 2023] Towards Data-Agnostic Pruning at Initialization: What Makes a Good Sparse Mask?

This is an official source code for "Towards Data-Agnostic Pruning at Initialization: What Makes a Good Sparse Mask?", NeurIPS 2023.

Requirements

Python >= 3.6, Pytorch >= 1.8, Cvxpy

How to Run

sh scripts/run.sh

References

This source code is adopted from "Pruning neural networks without any data by iteratively conserving synaptic flow".