Pedro Goncalves Mokarzel

Related comment in https://github.com/pytorch/xla/pull/9184/files#r2115084512

@benawilson will be taking a look at this issue.

@rpsilva-aws >> We had minimally prototyped it with XLA + SPMD - mimicking communication across heterogeneous SPMD worlds, minus the native distributed pipelining APIs. Is this prototype on github? I...

I would like to confirm my understanding of *Model initialization*: From my reading, you suggest that we change data loading in the case of SPMD+MPMD such that we only load...

From https://github.com/pytorch/xla/issues/9019#issuecomment-2938686214. We are looking at Ray vs IFRT. I am wanting to write an RFC on the topic as I think it is a subject worth diving deeper into....

When did these tests start failing? https://github.com/pytorch/xla/pull/8849 was merged March 26th. Also, could you explain a little more on what is causing the error? My understanding from the original PR...

> Is there a possibility to not make API changes but instead allow setting a flag for when we want PJRT to know about less than the global view? >...

I have been investigating how the specific implementation of this bug might look like. I scoured our code for where xla::OpSharding is being used, and here are my findings: Previously...

Hello, me and @GleasonK both took a look at this a while ago. I will add here what I know, and I will let Kevin correct me or add anything:...

I did not. I have done so, and it seems to be working properly now. I initially was following https://github.com/AI-Hypercomputer/xpk/tree/6ba0019ac14f9c0b1d5c51371bc40bc43e3b65a8?tab=readme-ov-file#installation, and just used `pip install xpk`. Should the instructions default...