Skip to content

Conversation

allenwang28
Copy link
Contributor

@allenwang28 allenwang28 commented Sep 1, 2025

Monarch doesn't support nested actors well yet, i.e. an actor that itself spins up processes and spawns actors on those processes. It will eventually support this, needs ProcMeshRef landed, which is actively being developed. As is, these nested actors aren't able to communicate with each other which is why we were seeing errors in torchstore.

This PR introduces a Service implementation that isn't based on an Actor, while retaining the exact same APIs. It's basically the same Service code as before, just removing the endpoints/calls/spawning a service actor etc.

The plan for this will be to remove the current implementation in favor of what we're calling "V2" whenever ProcMeshRef is well supported.

@meta-cla meta-cla bot added the CLA Signed This label is managed by the Meta Open Source bot. label Sep 1, 2025
Copy link
Contributor

@LucasLLC LucasLLC left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What a legend

Copy link
Member

@joecummings joecummings left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Approving to unblock but can you list out what the changes made are?

@allenwang28 allenwang28 changed the title create a service without nesting (don't land yet) Introduce a temporary Service implementation that isn't an Actor Sep 1, 2025
@allenwang28 allenwang28 merged commit 0a7ad3e into meta-pytorch:main Sep 1, 2025
4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed This label is managed by the Meta Open Source bot.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants