amirgholami / PyHessian

PyHessian is a Pytorch library for second-order based analysis and training of Neural Networks
MIT License
694 stars 119 forks source link

How to execute pyhessian using DDP. #26

Open MiracleLin001 opened 1 week ago

MiracleLin001 commented 1 week ago

I note that in the abstract of the paper it mentions it supports distributed-memory execution on cloud/supercomputer systems and is available as open source, however not a single actual example exists in the repo. Since the cost of hessian is significant, is it pratical to use DDP to reduce GPU memory load? Or any examples? Thank you.