Asher Trockman
Asher Trockman
Sure, that sounds good. I'll try to finish a working demo soon.
This still needs some testing, but it looks promising. In a brief experiment on CIFAR-10, this RandAugment implementation added +1% test accuracy and cost about 0.1s/epoch. The first epoch takes...
Thanks for pointing this out. I'll look into automatically adapting the image size. Did it work after fixing the image size in your example?
Great, thanks! Hopefully your experiments work out. It looks like the `size` argument was unnecessary, and it should now work even when changing the image size mid-training.
> This is awesome! When will it be merged into master / a release? Thanks! I'm not sure -- the plan was to merge it with release v1.0.0 (#160), but...
@andrewilyas @GuillaumeLeclerc Are you still interested in merging #154 into V1.0.0? For what it's worth: while I didn't finish the checklist in the PR, some colleagues and I have been...
Yeah, that sounds good to me. There are tests here: https://github.com/libffcv/ffcv/blob/0fa3d7c8a041fec7348025afafe6e4f6fff60f33/tests/test_rand_aug.py. My goal was to very closely replicate the RandAugment implementation in TorchVision, so most of the tests use it...
> Hi @ashertrockman ! It seems I lost track of this PR a while ago - do you think it's feasible to merge into v1.1.0? Yeah, I think it should...
It has been a while since I updated this for iOS 10, and I don't think I made any notes on the process, so I don't have specifics. I think...
No problem, and thanks! Good luck with your project.