Skip to content

[Question] General questions on optree and community use and support #220

@matthewfeickert

Description

@matthewfeickert

Required prerequisites

Questions

👋 @XuehaiPan We (members of the Scientific Python community) are quite excited to see optree as some of our members have been looking around at standalone PyTree implementations for a bit. It is great to see that optree has been in PyTorch since 2023 (via pytorch/pytorch#93139 motivated by pytorch/pytorch#65761 (comment)) and that there is integration with PyTorch, NumPy, and JAX. From some of our initial testing it seems like optree is very similar to the JAX implementation, though slower by a factor of 2 for very large PyTrees (you already mention that optree was heavily motivated by the JAX implementation, so this makes sense that it is similar).

Some general questions we had:

  • Have you made the JAX team aware of optree and discussed it with them?
  • Is the https://github.com/metaopt/ organization a research org for your own purposes? Or does it have a larger ecosystem / goal it is part of? (cc @Benjamin-eecs as they're listed as the email contact of the org)
  • Is there a roadmap for optree in general? Or is it "complete" and further feature development work is not planned?

cc @pfackeldey @mihaimaruseac (also for more technically informed comments than I have 😄)

Metadata

Metadata

Assignees

No one assigned

    Labels

    questionFurther information is requested

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions