Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

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

2.638Validation Loss

SF-AdamW

2.6365242.6464872.656452.666413Dec 18, 2025
Updated 4d ago

Evaluation Results

MethodLinks
2025.12
2.638
2025.12
2.638
2025.12
2.638
2025.12
2.638
2025.12
2.645
2025.12
2.6553
2025.12
2.6558
2025.12
2.6572
2025.12
2.6577
2025.12
2.6602
2025.12
2.6639
2025.12
2.6737
2025.12
2.6749
2025.12
2.6749
2025.12
2.6749
2025.12
2.6749