We have offered several ways to evaluate and train your model with VimGolf:

  1. VimGolf Terminal Bench adaptor (Evaluation only)
    Pull Request: terminal-bench#990
    Usage: Adaptor usageParity experiment reproduction

  2. VimGolf for Inspect Evals (Evaluation only)
    Pull Request: inspect_evals#517
    Usage: Basic usage

  3. VimGolf-Gym (Training & Evaluation)
    Usage: Training environmentBenchmark & evaluation

  4. VimGolf for Prime Intellect (Training & Evaluation)
    Usage: Basic usage

  5. VimGolf for SkyRL (Training & Evaluation)
    Pull Request: skyrl#272
    Usage: Basic usage

  6. VimGolf for RLLM (Training & Evaluation)
    Pull Request: rllm#209
    Usage: Basic usage