Shypd.aiThe largest open index of AI toolsPyHessian is a Pytorch library for second-order based analysis and training of Neural Networks. It enables computing metrics like top Hessian eigenvalues, trace of the Hessian matrix, and Eigenvalue Spectral Density.
Trending
About
PyHessian is a Pytorch library designed for advanced analysis and training of neural network models using second-order methods. It provides functionalities to compute critical metrics such as the top Hessian eigenvalues, the trace of the Hessian matrix, and the full Hessian Eigenvalues Spectral Density (ESD). This library is particularly useful for researchers and developers looking to gain deeper insights into the optimization landscape of neural networks, understand model behavior, and explore second-order optimization techniques. It supports both installation via pip and from source, and includes examples for training ResNet20 models on Cifar-10 and performing Hessian analysis on saved checkpoints.
Capabilities
Pricing & Plans
Open Source
Free
FAQs
Trending
