Google Gemma 2 Bootcamp Resources Released

title: [Google Gemma] Gemma2 bootcamp Public Resource Records
published: false
date: 2024-07-08 00:00:00 UTC
tags: 
canonical_url: https://www.evanlin.com/google-gemma2/

Preface:

Related Content

Overview

• Gemma2 Model Card

Gemma Cookbook Data Sharing

• Gemma Cookbook Github
  • You can contribute, but don't duplicate.
  • You can take a look at the wishlist
  • Chinese translation (by 保哥) https://github.com/doggy8088/gemma-cookbook

Try Gemma2 in AI Studio

Model Card

• Gemma1.1(2B, 7B)
• Gemma2 (27B, 9B), the small model 2.7B is still in training.

New: Gemma2 on Groq (If you don't have a good graphics card, you can use this example)

• Gemma2_on_Groq.ipynb

New: Gemma2 related Chinese models

• https://huggingface.co/cloudyu/google-gemma-7b-chinese-sft-v1 (Made by students from Tsinghua University, unofficial)

Q&A

• Q: Gemma2 - Fine-Tuning experience?
  • There are comments on HuggingFace, if you want to Fine-Tune, you need to change Attention to Egar
• Q: Does Gemma2 have an Embedding Model?
  • Not yet.
• Q: Is there a way to use Gemma directly through the cloud? No need to deploy GPU locally?
  • Currently there is a cookbook for Vertex AI on Gemma