Discover What's New In Gemma 1.1 Update: New 2B & 7B Instruction Tuned models

Рет қаралды 6,931

Күн бұрын

Google has released a set of new versions of their instruction tuned versions of Gemma. In this video I go through what they are good at and show new results on things like ReAct prompting.
Code Examples
Colab Gemma 1.1 7B: drp.li/K1OGY
Colab Gemma 1.0 7B: drp.li/BGJw3
Colab Gemma 1.1 2B: drp.li/vkuJ0
🕵️ Interested in building LLM Agents? Fill out the form below
Building LLM Agents Form: drp.li/dIMes
👨‍💻Github:
github.com/samwit/langchain-t...
github.com/samwit/llm-tutorials
⏱️Time Stamps:
00:00 Intro
00:25 Gemma 1.1 Hugging Face
01:21 Gemma 1.1 Release Notes
02:21 Google Developer Day in London Videos on KZfaq
02:47 Code Time
04:29 Prompt Format
04:44 Analogy Sample
09:08 GSM 8K Sample
10:13 A Meditation Mantra is All you Need
11:36 Gemma 1.1 Translation
13:14 ReAct Prompting
16:48 Gemma 2B 1.1 Testing
#gemma #localllms

Пікірлер: 18

@user-kx7tg7vt6n 2 ай бұрын

Hey, thanks for the video ! 🙂 I'm noticing some words cutting out now and then - I'm sure you're aware too just wanted to mention it ! Again, thanks for everything 🙂

@SonGoku-pc7jl Ай бұрын

very interesting, best video of gemma 1.1 ;)

@samwitteveenai Ай бұрын

thanks much appreciated

@theunknown2090 2 ай бұрын

Thanks for the awesome video

@andyma1146 2 ай бұрын

In the section “A meditation mantra is all you need” you got better performance by adjusting the system prompt. I’m not very good at prompt engineering. Do you think that I could get comparable performance improvements if instead of playing with the system prompt, I asked LLM to reflect on or revise its answer? This would be kind of like using an agentic workflow. I’m wondering if using reflection would allow me to avoid prompt engineering for each new model I want to try out 🧐🤨

@user-zq9bp5yv6z 2 ай бұрын

thanks for aharing🎉. a question about 13:03 . What do you mean by output things with "tags" as what people are doing with Claude3.?

@samwitteveenai 2 ай бұрын

Claude models use XML tags for outputting and for partitioning things like exemplars for FewShot Learning etc. I cover this in the video about Claude Haiku with examples.

@MeinDeutschkurs 2 ай бұрын

Can‘t you tell to send the „stop-token“ after the tool decision? 15:46

@samwitteveenai 2 ай бұрын

yes though you want to send in the tool output for the next call then you don't want it using the stop token there. You want it generating to the next tool or decision etc.

@MeinDeutschkurs 2 ай бұрын

@@samwitteveenai 🤦‍♂️, in deed. I just thought about: how to avoid a very verbose reply.