List your company

Language Modeling on LLaMa-2 1B (val)

2.421Validation LossIsoLoCo
2.419842.427672.43552.44333Jul 3, 2026

Evaluation results

MethodLinks
IsoLoCoModel size=1B, Number of workers (R)=8, Synchronization interval (H)=302026.072.421
DiLoCoModel size=1B, Number of workers (R)=8, Synchronization interval (H)=302026.072.45