Gwern Branwen

The API doesn't necessarily *have* to change. Perfect is the enemy of better: Space could be left in the AST type just as it is now, but deprecated and not...

I see, thanks. Perhaps you could update the README.md for this project with a note that it's obsoleted by animeface-2009 and that animeface-2009 also provides a script for extracting faces?

If you are unable to provide any kind of git clone or archive download due to lack of time, please at least don't go out of your way to detect...

It is your server blocking requests, and even if you did not explicitly enable it, that does users no good. How exactly are we supposed to use the anonymous repos...

I wouldn't because anonymized peer review is deeply flawed and has been repeatedly experimentally shown to be highly unreliable and largely useless compared to much better methodologies like pre-registered reports...

A reproachful tone is perfectly appropriate when dealing with a maintainer whose response to a serious bug is to deflect and give the most narrowly technically-correct (surely the best possible...

Further user-blaming and tone-policing to shift the blame to anyone but oneself. The bug remains.

I get the same error with Pytorch 1.0.0 (along with an OOM error on the VNet model, which seems odd given that saving memory is exactly what gradient checkpointing is...

Yes, I ran into the same issue with `` tags as well: ``` $ mdl --version 0.8.0 $ cat foo.page [link](http://www.example.com) $ mdl foo.page foo.page:2: MD034 Bare URL used A...

It's worth noting there's another TPU implementation where they claim to have trained models successfully: https://github.com/giannisdaras/smyrf/tree/master/examples/tpu_biggan (supporting code for ["SMYRF: Efficient attention using asymmetric clustering", Daras et al 2020](https://arxiv.org/abs/2010.05315)). Tensorfork...