Thank you so much for providing high level APIs to compute hessian matrix and jacobian matrix (Automatic differentiation package - torch.autograd — PyTorch 1.7.1 documentation). I also noticed toy examples provided.
However, there are a lot of posts trying to figure it out how to use it on complicated networks’ parameters (Search results for 'hessian' - PyTorch Forums, machine learning - How to compute hessian matrix for all parameters in a network in pytorch? - Stack Overflow, python - PyTorch most efficient Jacobian/Hessian calculation - Stack Overflow). If possible, would you please release some examples of using it with descent networks (with forward and get_loss)?