invoiceguard-code / training /train_sft.py

Commit History

  1. v5d: save best-epoch checkpoint during SFT

    Verified signature
    a674764
  2. v5c: add --min-investigation-steps filter for submit-only with longer contexts

    Verified signature
    2857a20
  3. v5b: add --submit-only SFT mode

    Verified signature
    2a667e7
  4. v5: remove 4-bit quant, variable-length traces, submit oversampling, GRPO exploration fixes

    Verified signature
    1d75102
  5. Upload folder using huggingface_hub

    Verified signature
    4cd2128
  6. Fix Qwen3 thinking mode + increase max_new_tokens: training/train_sft.py

    Verified signature
    7c48dcf
  7. Add InvoiceGuard SFT and merge tooling

    Verified signature
    25ee43d