fix(npu): bracket IPv6 hosts in distributed init methods#2199
Open
EazyReal wants to merge 1 commit into
Open
Conversation
The documented NPU setup checks out slime v0.2.2 and applies docker/npu_patch/slime.patch. That pinned code still constructs three torch.distributed tcp URLs from bare node addresses, so IPv6 hosts produce ambiguous init methods. Extend the NPU patch to import the existing _wrap_ipv6 helper and use it in the Megatron actor/critic, distributed weight-update, and FSDP weight-update paths. The resulting patch applies cleanly to v0.2.2, leaves no unwrapped target URLs, and the three patched modules compile.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
PR #2151 fixed the current-main weight-update rendezvous URL, but the documented NPU setup checks out Slime v0.2.2 and applies docker/npu_patch/slime.patch. That pinned code has three additional actor/critic and weight-update tcp init methods that still interpolate bare IPv6 addresses.
Extend the NPU patch to use the existing _wrap_ipv6 helper at those three sites. The updated patch applies cleanly to v0.2.2, leaves 3 wrapped and 0 unwrapped target URLs, and all three resulting modules compile.