the math and in that math, it's optimizations, mathematical operation optimizations, and then how do you get it into the hardware?

Right, exactly.

It's not like, oh, I can take forever to find the optimum, right? Like we all have things to do and we want to get answers.

And so you're trying to find the fastest way to get to the best approximation of the maximum or minimum, right? that you can get.

And so there's different parts of the PyTorch ecosystem.

There's other pieces that are dealing with actually hard compiling and minimizing the size of the weights and optimizing the speed of the weights, maybe the sacrifice for being on device.

Right, so there's a big effort within the PyTorch community, particularly from the embedded manufacturers to get a version of PyTorch or to allow PyTorch to produce an asset that's smaller and optimized for our phone or on a sensor sitting off somewhere and still be able to run it and get the pieces out of it that you want, right? So there's just, there's a lot going on.

What are some of the other key projects that are important to highlight in the PyTorch Foundation?