The "Pytorch/XLA overview" is very long, goes into advanced topics, and is overall intimidating for new users.

[yaoshiang]

📚 Documentation

The "Pytorch/XLA overview" includes many advanced topics that go beyond an "overview", including how to specifically convert Stable Diffusion to run on TPUs (which is more of a Guide) and how to profile (which is more of a Tutorial). The result is an intimidating introduction for potential users of PyTorch/XLA.

I'd suggest we break the SD section into a stand-alone guide. And the profiling section into a standalone tutorial, one with a simplified example that has a successful outcome (the current section ends with a "we found the problems but there's nothing we can do").

The remaining copy can be redrafted into an "intro", so that users can hear about some of the benefits of PyTorch/XLA and as a result get encouraged to continue reading and even trying out the platform.