Hugging Face's logo

Spaces:
testingaccc
/
conflict-arbitration-env
Sleeping

App Files Community
conflict-arbitration-env / training

Commit History

  1. ship submission: judge-aligned README, real training curves from 2000-step run, blog, scripts⁠

    Jeeevan11
    ce00c50

  1. optimize for hackathon time budget: 256 tokens, 200-step checkpoints

    Jeeevan11
    bd2780e

  2. normalize agent C decision before sending to env (defaults + clamps)

    Jeeevan11
    2159f38

  3. disable unsloth fast_inference (vllm not installed)

    Jeeevan11
    2594ad1

  4. bump transformers to >=4.50.3 (unsloth needs Qwen3 support)

    Jeeevan11
    966f3d2

  5. skip flash-attn build; install unsloth + deps manually

    Jeeevan11
    68f46e8

  6. upgrade torch to 2.5+ for unsloth_zoo compatibility; use ampere-torch250 wheel

    Jeeevan11
    e1ae66a

  7. import unsloth first; pin torch 2.4.0 for unsloth cpp ext compat

    Jeeevan11
    569dcfa

  8. broaden TRL import catch; pin compatible transformers/trl/peft versions

    Jeeevan11
    69a91cc

  9. entrypoint: skip clone, expect code in cwd

    Jeeevan11
    fd75dec

  10. add training driver, job entrypoint, seed + upload support

    Jeeevan11
    68ec12d

  11. add training driver + windows setup guide

    Jeeevan11
    ce42707

  12. expand dataset to 40 templates / 8 domains; fix injection edge cases; lazy unsloth import

    Jeeevan11
    b532a41

  13. expand dataset: 40 templates, 8 domains, fixed injection edge cases

    Jeeevan11
    4a18ff6

  14. initial commit: conflict arbitration env

    Jeeevan11
    8fd5450