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

fix: Improve partitioning + lowering systems in torch.compile path #1879

Merged
merged 2 commits into from
May 19, 2023

Conversation

gs-olive
Copy link
Collaborator

@gs-olive gs-olive commented May 2, 2023

Description

  • Improve torch.compile Dynamo partitioning system by incorporating key arguments including min_block_size and torch_executed_ops, which are available for use in TorchScript
  • Improve torch.compile lowering system by adding key new decompositions to improve coverage and reduce the number of unique operators requiring implementation
  • Update testing framework to use utilities, reducing code replication
  • Add extensive testing of new partitioning system and lowering phases

Note: Not to be merged prior to release/1.4 branch finalization (features are experimental and evolving)

Fixes #1832

Type of change

  • Bug fix (non-breaking change which fixes an issue)

Checklist:

  • [ x ] My code follows the style guidelines of this project (You can use the linters)
  • [ x ] I have performed a self-review of my own code
  • [ x ] I have commented my code, particularly in hard-to-understand areas and hacks
  • [ x ] I have made corresponding changes to the documentation
  • [ x ] I have added tests to verify my fix or my feature
  • [ x ] New and existing unit tests pass locally with my changes
  • [ x ] I have added the relevant labels to my PR in so that relevant reviewers are notified

@gs-olive gs-olive self-assigned this May 2, 2023
@github-actions github-actions bot added the component: api [Python] Issues re: Python API label May 2, 2023
@github-actions github-actions bot requested a review from narendasan May 2, 2023 00:06
@gs-olive gs-olive added the component: dynamo Issues relating to the `torch.compile` or `torch._dynamo.export` paths label May 2, 2023
@gs-olive gs-olive force-pushed the dynamo_partitoner_lowering branch 3 times, most recently from f745135 to fc63b2c Compare May 2, 2023 03:55
Copy link
Collaborator

@narendasan narendasan left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@gs-olive gs-olive changed the base branch from dynamo_changes to main May 3, 2023 00:14
@gs-olive gs-olive force-pushed the dynamo_partitoner_lowering branch from fc63b2c to 0d85466 Compare May 3, 2023 00:55
@gs-olive gs-olive force-pushed the dynamo_partitoner_lowering branch 3 times, most recently from c3f88c1 to a77017c Compare May 12, 2023 22:21
- Improve torch.compile Dynamo partitioning system by incorporating key
arguments including `min_block_size` and `torch_executed_ops`, which are
available for use in TorchScript
- Improve torch.compile lowering system by adding key new decompositions
to improve coverage and reduce the number of unique operators requiring
implementation
- Update testing framework to use utilities, reducing code replication
- Add extensive testing of new partitioning system and lowering phases
@gs-olive gs-olive force-pushed the dynamo_partitoner_lowering branch 2 times, most recently from 4f527a0 to 5d77b6b Compare May 18, 2023 20:25
- Update default `min_block_size`
@gs-olive gs-olive force-pushed the dynamo_partitoner_lowering branch from 5d77b6b to 357d18d Compare May 18, 2023 21:40
@gs-olive gs-olive merged commit 98376b3 into pytorch:main May 19, 2023
@gs-olive gs-olive deleted the dynamo_partitoner_lowering branch May 19, 2023 01:24
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: 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.

Improve torch.compile partitioning system and adapt lowering passes from TorchScript to Dynamo
3 participants