New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Gemma: Open Models Based on Gemini Research and Technology, 2024 #1277

Open

AkihikoWatanabe opened this issue Apr 8, 2024 · 2 comments

Labels

LanguageModel NLP OpenWeightLLM

Owner

AkihikoWatanabe commented Apr 8, 2024 •

edited

Loading

https://storage.googleapis.com/deepmind-media/gemma/gemma-report.pdf

blog: https://blog.google/technology/developers/gemma-open-models/

AkihikoWatanabe added NLP LanguageModel OpenWeightLLM labels

Owner Author

AkihikoWatanabe commented May 24, 2024 •

edited

Loading

アーキテクチャはTransformer Decoderを利用。モデルのサイズは2Bと7B。
オリジナルのTransformer Decoderアーキテクチャから、下記改善を実施している：

Multi Query Attention Fast Transformer Decoding: One Write-Head is All You Need, Noam Shazeer, N/A, arXiv'19 #1272 を利用
RoPE Embedding RoFormer: Enhanced Transformer with Rotary Position Embedding, Jianlin Su+, N/A, Neurocomputing, 2024 #1310 を利用
GeGLU GLU Variants Improve Transformer, Noam Shazeer, N/A, arXiv'20 #1311 の利用
RMSNormの利用（学習を安定させるため; LLaMAと同様）

Owner Author

AkihikoWatanabe commented May 24, 2024 •

edited

Loading

Mistral #1309 よりも高い性能を示している：

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment