invoiceguard-code / training /train_sft.py

Commit History

v5d: save best-epoch checkpoint during SFT
a674764
verified

piyush-mk commited on

v5c: add --min-investigation-steps filter for submit-only with longer contexts
2857a20
verified

piyush-mk commited on

v5b: add --submit-only SFT mode
2a667e7
verified

piyush-mk commited on

v5: remove 4-bit quant, variable-length traces, submit oversampling, GRPO exploration fixes
1d75102
verified

piyush-mk commited on

Upload folder using huggingface_hub
4cd2128
verified

piyush-mk commited on

Fix Qwen3 thinking mode + increase max_new_tokens: training/train_sft.py
7c48dcf
verified

piyush-mk commited on

Add InvoiceGuard SFT and merge tooling
25ee43d
verified

piyush-mk commited on