Explore quantizing the curvature matrices to reduce their memory footprint when running inference.
Explore quantizing the curvature matrices to reduce their memory footprint when running inference.