cannon-and-wall / training

Commit History

  1. fix: Gradio mount, verifier stage3, poc_history clear, real notebook

    5844c1d

  1. rewrite: training notebook - drop Unsloth, use transformers+peft+bitsandbytes

    8cba85d
  2. feat: GRPO training, fixed curriculum stages 2+3, AST verifier, leaderboard UI, openenv schema

    da648a3
  3. fix: remove 3 debug cells from notebook

    592bdde
  4. feat: add real training notebook from Colab run

    fd58a6d
  5. initial deploy — jairaj files, teammate files pending

    9556146