List your company

Language Model Pre-training on C4 Llama-1B scratch (val)

2.638Validation LossSF-AdamW
2.6365242.6464872.656452.666413Dec 18, 2025

Evaluation results

MethodLinks
SF-AdamWNumber of inner steps (H)=162025.122.638
SF-AdamWNumber of inner steps (H)=322025.122.638
SF-AdamWNumber of inner steps (H)=642025.122.638
SF-AdamWNumber of inner steps (H)=1282025.122.638
GPA-AdamWNumber of inner steps (H)=642025.122.645
GPA-AdamWNumber of inner steps (H)=1282025.122.6553
DiLoCo-AdamWNumber of inner steps (H)=1282025.122.6558
DiLoCo-AdamWNumber of inner steps (H)=642025.122.6572
DiLoCo-AdamWNumber of inner steps (H)=322025.122.6577
GPA-AdamWNumber of inner steps (H)=322025.122.6602
GPA-AdamWNumber of inner steps (H)=162025.122.6639
DiLoCo-AdamWNumber of inner steps (H)=162025.122.6737
AdamWNumber of inner steps (H)=162025.122.6749
AdamWNumber of inner steps (H)=322025.122.6749
AdamWNumber of inner steps (H)=642025.122.6749
AdamWNumber of inner steps (H)=1282025.122.6749