Birch-san

Okay it's been pointed out that this is already a to-do in the readme 😛

if I'm understanding correctly, then yeah @marunine experimented with sampling approaches, including dpm-solver. hopefully m9 can explain — I don't wanna misrepresent the conclusions.

hmm but this problem looks different to the category of problem tracked by the linked issue. @Maxhirez I'd encourage you to submit this as an issue to [pytorch](https://github.com/pytorch/pytorch/issues), especially as...

> is it more likely to be with the nightly build of torch or the CLIP model trying to use it? mm my guess would be torch. the dream of...

Webpack config is documented here:. https://github.com/Birch-san/box2d-wasm/issues/26

Actually if it's a syntax error that sounds like a different problem: parsing ES6 as CommonJS

no problem; glad it's appreciated. here's some extra diffs that illustrate what's included in the rebase: https://github.com/google/liquidfun/issues/81#issuecomment-787498634 > except for Testbed okay, yeah that makes sense… `draw.cpp` needs to implement...

@pcuenca wait, you got it running on-GPU? and it was faster? that's massively different from the result I got. here's how I made it run on MPS: https://github.com/Birch-san/min-dalle/compare/Birch-san:min-dalle:main...Birch-san:min-dalle:mps _there's other...

> generation did not work properly it's worth knowing that the MPS backend _does_ have some silent errors where it will produce incorrect output (or at least transfer the wrong...

> I was also looking into getting this model on the phone. Apple says that for transformers in pytorch, the dimensions aren't in optimal order for the neural engine: https://machinelearning.apple.com/research/neural-engine-transformers...