Skip to content

feat: Gymnasium-style environment adapter #1030

@cwing-nvidia

Description

@cwing-nvidia

Use cases, pain points, and background
Many researchers and engineers in the RL community are very familiar with the Gymnasium API: reset(), step(), reward, done. When they encounter NeMo Gym's environment architecture - FastAPI servers, HTTP communication, Hydra YAML configs, resources servers, agent servers - there's a new learning curve. The concepts don't map immediately to what they already know.

Description:
Provide a Gymnasium-style environment adapter that lets users define environments using familiar APIs

Design:
What files should be touched? What logic should be written?

Out of scope:
What are some items that this issue could be mistaken to cover that this issue should explicitly NOT cover?

Acceptance Criteria:

  • Individual items that need to be finished in order for this issue to be considered completed

Metadata

Metadata

Assignees

No one assigned

    Labels

    core-infraHelpful infrastructure

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions