Skip to content

Add spmd_backend config flag - #3540

Closed
pianpwk wants to merge 1 commit into
gh/pianpwk/26/basefrom
gh/pianpwk/26/head
Closed

Add spmd_backend config flag#3540
pianpwk wants to merge 1 commit into
gh/pianpwk/26/basefrom
gh/pianpwk/26/head

Conversation

@pianpwk

@pianpwk pianpwk commented Jun 4, 2026

Copy link
Copy Markdown
Contributor

Stack from ghstack (oldest at bottom):

Rename parallelism.full_dtensor: bool to parallelism.spmd_backend: Literal["default", "full_dtensor", "spmd_types"] and add debug.spmd_typechecking for the upcoming local-tensor SPMD path.

All existing callsites are updated mechanically:

  • full_dtensor=True/False -> spmd_backend="full_dtensor"/"default"

  • self.full_dtensor -> self.spmd_backend == "full_dtensor"

No new SPMD mesh construction or runtime plumbing yet. The spmd_types backend value is accepted here, and mesh/runtime logic lands in the next commit.

[ghstack-poisoned]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ciflow/8gpu 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.

1 participant