d-li14/PSConv

Optimized version of PSConv

rosinality opened this issue · 3 comments

rosinality commented

Hello, thank you for nice work!

It seems like that you made optimized version of PSConv as mentioned in the supplementary material of the paper. As it shows a great speedup, it could be very useful implementations for applying PSConv to various models.

Do you have a plan to release this optimized version of PSConv?

d-li14 commented

Thanks for your interest! Actually it is optimized by the Intel hardware-related team, they may keep it as confidential for future Intel AI HW design. In short, I'm sorry to say that we don't have a recent plan for the speedup version release.

rosinality commented

That's too bad. Anyway, thank you for your work! I think it is a great progress for multiscale architectures.

d-li14 commented

I have no access to this part except for the optimized statistical results. Probably seek support from @YaoAnbang.