Support prompt-completion data for LoRA fine-tuning#1423

Open

SahilKumar75 wants to merge 1 commit into

ml-explore:mainfrom

SahilKumar75:feature/lora-instruct-tuning

SahilKumar75 commented May 13, 2026

Summary

support prompt-completion JSONL records in the LoRA data loader
mask prompt tokens so loss is computed on completion tokens only
document the instruction tuning data format

Testing

python3 -m pytest lora/test_lora.py -q
python3 -m black --check lora/lora.py lora/test_lora.py
python3 -m pre_commit run --files lora/lora.py lora/test_lora.py lora/README.md

Closes #484


          Support prompt completion LoRA data

0d863ab

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet