Gemma 2: Improving Open Language Models at a Practical Size
paper
Fields
id
sources/model-fr-gemma-2-reportPublisher
Google DeepMind
Kind
paper
Published
2024-08-27
Accessed
2026-10-03
notes
Technical report describing the alternating local/global attention, logit softcapping for numerical stability at long context, and knowledge distillation from the larger Gemma 2 teacher.
Notes
Technical report describing the alternating local/global attention, logit softcapping for numerical stability at long context, and knowledge distillation from the larger Gemma 2 teacher.
Cross-references
This record references no other records.