QLoRA
ArchiveMedium
QLoRA fine-tunes large models on one GPU: quantize the base model to _____(1), train LoRA adapters in _____(2).
Flag format: CONGRESS{1:[precision],2:[precision]}
Example: CONGRESS{1:8bit,2:half_precision}
Show hint
Freeze and compress the big model, train tiny adapters at full quality.
Archive — no submissions accepted
This challenge is preserved for reference. Play live challenges at /challenges.