Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[GraphBolt][CUDA] Cooperative Minibatching initial exchange. #7795

Merged
merged 5 commits into from
Sep 11, 2024

Conversation

mfbalin
Copy link
Collaborator

@mfbalin mfbalin commented Sep 11, 2024

Description

Towards implementing #7273.

The code is untested right now. I need to merge some of the changes so that the PR size stays limited.

The initial exchange is required because each process needs to sample for the nodes they are assigned. There is no guarantee about any partitioning after things come from ItemSampler or when negative edges are added. Thus, initial exchange ensures that each process samples only for the nodes that they own.

Follow up work:

  1. Exchanges after each sampling stage.
  2. Add torch layer for GNN forward backward using Cooperative Minibatching and stored tensors including exchange information.
  3. Add multi-GPU example showcasing Cooperative Minibatching.
  4. Optimize Cooperative Minibatching using partitioning. (More details to come later.)

Checklist

Please feel free to remove inapplicable items for your PR.

  • The PR title starts with [$CATEGORY] (such as [NN], [Model], [Doc], [Feature]])
  • I've leverage the tools to beautify the python and c++ code.
  • The PR is complete and small, read the Google eng practice (CL equals to PR) to understand more about small PR. In DGL, we consider PRs with less than 200 lines of core code change are small (example, test and documentation could be exempted).
  • All changes have test coverage
  • Code is well-documented
  • To the best of my knowledge, examples are either not affected by this change, or have been fixed to be compatible with this change
  • Related issue is referred in this PR
  • If the PR is for a new model/paper, I've updated the example index here.

Changes

Sorry, something went wrong.

@mfbalin mfbalin added the expedited if it doesn't affect the main path approve first to unblock related projects, and review later label Sep 11, 2024
@mfbalin mfbalin requested a review from frozenbugs September 11, 2024 19:32
@dgl-bot
Copy link
Collaborator

dgl-bot commented Sep 11, 2024

To trigger regression tests:

  • @dgl-bot run [instance-type] [which tests] [compare-with-branch];
    For example: @dgl-bot run g4dn.4xlarge all dmlc/master or @dgl-bot run c5.9xlarge kernel,api dmlc/master

@dgl-bot
Copy link
Collaborator

dgl-bot commented Sep 11, 2024

Commit ID: 7ca6a7a

Build ID: 1

Status: ❌ CI test failed in Stage [Lint Check].

Report path: link

Full logs path: link

@dgl-bot
Copy link
Collaborator

dgl-bot commented Sep 11, 2024

Commit ID: c0af5c8

Build ID: 2

Status: ⚪️ CI test cancelled due to overrun.

Report path: link

Full logs path: link

@dgl-bot
Copy link
Collaborator

dgl-bot commented Sep 11, 2024

Commit ID: ba3bc97

Build ID: 3

Status: ✅ CI test succeeded.

Report path: link

Full logs path: link

@dgl-bot
Copy link
Collaborator

dgl-bot commented Sep 11, 2024

Commit ID: 9be2085

Build ID: 4

Status: ✅ CI test succeeded.

Report path: link

Full logs path: link

@mfbalin mfbalin merged commit 53e70c5 into dmlc:master Sep 11, 2024
2 checks passed
@mfbalin mfbalin deleted the gb_cuda_cooperative_exchange branch September 11, 2024 21:46
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
expedited if it doesn't affect the main path approve first to unblock related projects, and review later
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

2 participants