James Pirruccello

Shoot, I saw this after deploying to production. Can we change it to "fixed"?

I would be happy to merge a pull request that enables validation and also works well with unicode. However, I will not merge this pull request because it removes validation...

The problem is that Instagram silently ignores various characters that I would never expect to be ignored. Perhaps a better solution in my fork would be to create verbose errors...

Thanks @shijianjian . In my mind, this class would be similar in many respects to `kornia.augmentation.RandomResizedCrop`. If the current code for `RandomResizedCrop` looks like this: ```py class kornia.augmentation.RandomResizedCrop( size, scale=(0.08,...

I took a look but wasn't able to figure out how to safely move [ResizedCropGenerator](https://github.com/kornia/kornia/blob/3c747ccc0b79d726e8d7c98b3c38e2a6c40b358b/kornia/augmentation/_2d/geometric/resized_crop.py#L76) out of the `__init__`.

From what I can tell in the documentation, apex is still needed to take advantage of AMP with this library. Is that the case or is this issue completed?

For what it's worth, I get this error nondeterministically. When using faulthandler to try to localize it, I see the following: ``` Current thread 0x000000020b9b3a80 (most recent call first): File...

A bit of a late follow-up question, but I imagine that for traits that are near-perfect inverses of one another this will not be the case, and the sumstats need...

> However, if you really want to retain those multi-allelic variants, there is a way to do so. You can try modifying your local version of the MTAG code by...

Without commenting on what the order of operations *should always be*, the [implementation in this repo](https://github.com/sdoria/SimpleSelfAttention/blob/master/xresnet.py#L114) is consistent with: Conv2D -> Self Attention -> Activation