Hanxi Lin issues

Repositories
Issues
Comments

Results 2 issues of


                                            Hanxi Lin

Confused with the codes using grads clipping and accumulation simultaneously

Take args.iter_size==2 for example, I think the clipped and accumulated grads of your codes are clip(clip(grads1)+grads2), not clip(grads1+grads2), which makes more sense for me. I haven't run the code yet,...

Could you provide the pretrained model on SS-V1?

Some papers only report performance on SS-V1. It would be more convenient for comparing RubiksNet with these methods if the pretrained weights on SS-V1 are accessible.