Default QLoRA Learning Rate Too High for Small Datasets
Pretty-Ad774 · reddit · 2026-07-16
The author argues that the widely default 2e-4 learning rate in QLoRA tutorials can be a trap for small datasets, especially those with under 10,000 samples.
His practical observations:
- Many tutorials use 2e-4, but this originates from Alpaca's 52k sample size.
- On 5k–10k custom annotated datasets, models easily overfit in the first epoch.
- Training loss continues to drop, but validation loss stagnates or worsens.
- By lowering the learning rate to 1e-4 and increasing epochs from 3 to 5, validation performance improved significantly.
His conclusions:
- For over 30k samples, 2e-4 probably still works.
- For under 10k samples, start at 1e-4 or lower and carefully tune epochs.
- For sizes in between, don't just copy the default; actually tune the parameters.
He also expressed interest in reading any solid research that supports strictly using 2e-4 for small datasets.
More from Research
- Chinese AI labs are now treating distillation obfuscation as the top research topic — pmddomingos · 2026-07-22
- Structural ensembles beat single predictions in TCR:pMHC generalization study — quaidmorris · 2026-07-22
- RSS launches under OMSF to push structural biology data modeling at scale — MoAlQuraishi · 2026-07-22
- enFoldX tops 8 neoantigen scans and an unseen-peptide benchmark — quaidmorris · 2026-07-22
- enFoldX reaches AUC 0.82 on human VDJdb and transfers to mouse at 0.76 — quaidmorris · 2026-07-22
- enFoldX gains accuracy as AF3 ensemble disagreement rises for non-binders — quaidmorris · 2026-07-22