feat: add graph splitter for all-gather/all-reduce operations #1545
Conversation
This file contains 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
polvalente
changed the base branch from
main
to
pv-feat/experimental-sharding-backend
polvalente
changed the title
polvalente
merged commit 6eb6fba
into
pv-feat/experimental-sharding-backend
8 checks passed
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
For compilation passes we'll have:
ShardingPropagation -> Part of what's present in ShardingCompiler currently -- basically, this is the purpose for
__compile__. This is what propagates the sharding throughout the function.GraphSplitter -> This PR (see below)
ShardingSeparation -> This is also present in ShardingCompiler, and is mostly what
__jit__currently executes.This PR adds the GraphSplitter pass that we can use to split computations based on
certain conditions (such as operations that require that a given dimension is not sharded).
It also builds out a list of computation specs that we can use to build an execution graph afterwards.
Each of these stages might have further fan-out depending on the subsequent ShardingSeparation pass.