Skip to content

Skip adjacency partitioning when the graph is fully supported - #4530

Open
micwill755 wants to merge 1 commit into
mainfrom
partition-short-circuit
Open

Skip adjacency partitioning when the graph is fully supported#4530
micwill755 wants to merge 1 commit into
mainfrom
partition-short-circuit

Conversation

@micwill755

Copy link
Copy Markdown
Collaborator

Summary

  • Skip adjacency partitioning when require_full_compilation already has full converter coverage, wrapping the graph as a single TRT block instead of repeating AccNodesFinder, fusion, and split.

PR 1 of 3 in a compile-startup stack.

Test plan

  • Compile a fully supported model with require_full_compilation=True and confirm one _run_on_acc_* submodule
  • Compile a mixed-support graph and confirm adjacency partitioning still runs
  • Flux / VLA compile still succeeds with this path enabled

When require_full_compilation already has full converter coverage, wrap the graph as one TRT block instead of repeating AccNodesFinder, fusion, and adjacency split.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

cla signed component: api [Python] Issues re: Python API component: core Issues re: The core compiler component: dynamo Issues relating to the `torch.compile` or `torch._dynamo.export` paths

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant