pytorch
Document on how to profile with torch_xla
📚 Documentation
I found we don't have a doc/guide on how to profile with torch_xla. We should add this because getting profile is essential for performance analysis.
Yes, absolutely. Who can I work with to gather the information needed?
I have not heard back from anyone on this. I'll need someone to contact me to get me the information you would like added to the documentation. Thanks!
There is a profiling guide hidden as a section of the "intro", here.
https://github.com/pytorch/xla/blob/master/docs/source/learn/xla-overview.md
It was my intention to break this out into our own profiling guide, however, there is another profiling guide on cloud.google.com, so there is more work to think through this properly get it right with xprof, etc. @chrisjones is now leading all this. (I don't have his github handle).
Reached out to @chrisjones, will update when I have more information. In the meantime are there torch/xla specific profiling topics we want to address?