List your company

Ranking Evaluation on English fine-tuning corpus 3B tokens

0.904Observed LossLinear-H
0.903360.907680.9120.91632Sep 7, 2024

Evaluation results

MethodLinks
Linear-HSchedule=Linear-H, ηmax=1.5×10−3, a1=128, a3=1282024.090.9042.10211
Linear-LSchedule=Linear-L, ηmax=10−3, a1=512, a3=5122024.090.9092.15822
Linear-SSchedule=Linear-S, ηmax=5×10−4, a1=128, a3=1282024.090.922.30233