Kamer Ali Yuksel

FYI, I have just opened a feature request in Pytorch repository as well: https://github.com/pytorch/pytorch/issues/52626

Sounds great!!! If we could also use it with any given optimizer, that would be perfect. Nobody uses default Adam anymore but e.g. optimizers from torch_optimizer package.

@zhuchen03 I see that it requires dataloader and seems to be specific to the classification. This is too specific unfortunately for me to use in some novel problems that I...

@zhuchen03 In my case, it is a generative model that is trained by QHAdam (with an adaptive gradient clipping wrapper), which learns to continously generates population of solutions to e.g....

Thanks @zhuchen03, how can I send you a sample code? Can I use the (cs.umd.edu) e-mail that is mentioned at your resume?

A very similar question has been asked almost a year ago in this link: https://discuss.pytorch.org/t/share-gradients-in-multiprocessing/15181

Here is an implementation for Adaptive Gradient Clipping: https://github.com/vballoli/nfnets-pytorch/blob/main/nfnets/agc.py

Thanks, Yue! I would also strongly suggest including the following proximity-based method: https://github.com/UCSC-REAL/SimiFeat It is a state-of-the-art outlier detection method in presence of labels, based on k-NN queries.

Yes, that would be a great feature, even if that means not rendering whole batch but a selected one. If we could render all of them in a single scene...

I've written a loss function here for maximizing the eigenvalue entropy or gini-index of a portfolio. Also, contains a PyTorch implementation of the probabilistic sharpe ratio. Let me know if...