Llama 3 8B: BIG Step for Local AI Agents! – Full Tutorial (Build Your Own Tools)
AI技術の魅力は、自己学習や自律性によって人間の手を離れても進化し続ける能力にあります。今回紹介する「Llama 3」は、ローカルAIエージェントを構築する際に大きな進歩をもたらすツールです。
この記事では、「Llama 3」を使用してAPIや内部関数を利用してツールを操作するAIエージェントを作成する方法を詳しく解説しています。ローカルなllms（Long-Short Term Memory）である「Llama 3」を活用することで、高度なカスタマイズが可能であり、非常に強力な機能が実現できます。
具体的には、動画内では「Llama 3」を使用してツールや関数を呼び出す方法について詳細に解説されています。このような機能呼び出しは非常に重要であり、効率的かつ柔軟なプログラミングが可能となります。
さらに、記事では「Llama 3」エージェントのテスト方法や新しいツール／関数の作成方法も示されており、実践的な手順がわかりやすく紹介されています。これによって、「Llama 3」のパワーと多様性が理解しやすくなっています。
最後に、記事は全体のまとめとして結論部分が記されており、読者が得た知識や情報の整理が行われています。これによって記事全体の流れが締めくくられ、読者は内容をスムーズに理解できる工夫がされています。
AI技術の魅力的な点は、その持つ無限の可能性と進化性です。本記事ではその一端を垣間見ることができるため、AI技術への興味や理解を深めたい方々におすすめです。
can you use ollama create and make this a model for home assistant ?
so please tell me it can use info from the vault offline ,
Im sorry, good content, but you overexplain things and saying 'kind of' too much really gets to me
Hi. I tested your 'search_google(query)' function with the same query that you used in your demo and (not surprisingly) it returned somewhat different results. Also, in order to allow the fcn 'GoogleSearch' to execute, I found that I had to: 'pip install google-search-results' as well as 'serpapi'."
Thank you again for this very thorough and informative review!
This might sound like a really dumb question, but how would one approach this when using Jan and the interface? Is that at all possible?
Incredible tutorial. I have realized that I have never seen a full tutorial that is full like this one
Cheers
Thank you for sharing, where can I see the full code?
🎯 Key points for quick navigation:
00:14 📊 The agent uses search Google to collect information and put it into a rag (repository of articles and guides&t=14).
00:53 💡 Llama 3 was trained on up to 15 trillion tokens.
01:07 📨 The agent has a tool that allows it to send mail, and it can be used to send information to an email address.
02:01 📝 The system has several functions, including send email, search Google, check context, and chat.
03:39 🔍 The search Google function is triggered by user input containing "search Google" or similar queries.
04:19 💡 The model uses the dolphin Tre version of Llama 3 with a temperature set to zero for terministic responses.
05:37 🔮 The intelligent part of the system listens to user requests and prepares function calls based on those requests.
06:48 📝 The system translates secret instruction notes into simple readable dictionaries that tell the system which function to run and what information to use as arguments.
10:03 📊 The chat function returns a message content that can be parsed to understand the context of the conversation.
10:45 🔍 The system uses wrapper tags to identify specific functions, such as sending an email.
11:39 📧 The system can search for available AMA models and return relevant information.
12:22 💡 The system can perform a rag search to retrieve information and send it to the user's email address.
13:16 📧 The system can send emails successfully, making it a big step for local AI agents.
14:27 📝 A new function called "Write to notes" is added to integrate with the text file "notes.txt".
15:34 🔧 The "Write to notes" function is set up in the chat function with an if statement.
Made with HARPA AI
could you show us how we can use the Modular Accelerator Executable for reduction in latency of models?
Where can I download the code from this video?
at first. btw I'm still at first. it's cos I'm dumb. but have you code on lamma 3 a python file to make it be able to create a google search engine? like who doesn't have google search tool? I don't see the point!
i think you can try phi3 i found it fast and for instruct porpuses works fine
Hello… good video… thanks. Where can I find the Github-Link? i m member 🙂
I would do these things too, but scraping is against googles term of use and almost every site has something against scraping. Could potentially be risky for you
great video, but please remove annoying background music
Outstanding demonstration, thank you. I would like to make it possible to add new tools using a yaml or json configuration file so you don't need to hard-code anything. It would also be nice to have the answer returned immediately instead of having to prompt it to check the retrieved context. But all in all this one of the best and most practical open source AI demos I've seen.
Great video, but where can I find the code (lm3.py)?
Hi, I just joined, how can I get the git for this code?
How does it compare to the likes of Dolphin-Mistral which has the bigger context window/code training?
I paid to become a member to get the github access and the link works now – it was down for a bit, thanks to AAA for sorting it out
I've just joined the community. I'd like to use LM Studio instead of ollama as ollama doesn't like windows. how can I adapt the local rag with LM studio, and I would like to create some agents. Thank you.
This is what always gets me about function calling, unless you intend to make the endpoint available to the public, it would be far easier and faster to just expose the function call. So instead of asking the AI to search, you just call the search command directly. Using AI to figureout what function to call and the needed parameters, is like using a flame thrower to light a candle. It only makes sense if the user doesn't understand function calls.
Great Content, I joined the members community but still no code for this one on GitHub, Please share how can we access the code. Thanks!
I cant host 70b locally or even 8b, i need to use an llama3 hosted api with agents that crawl many urls, can you suggest a setup that would work for that
Amazing content. What of your videos can I watch prior to these steps to understand how to get Ollama?.
new Argy sub.
Great video! Thank you very much for the explanation.
Do you know hardware requirements for 70b model (locally)?
Llama 3 8B instruct wipe the floor with every 70B Llama 2 model,its a freaking beast with massive knowledge about anything you put on it.I try to dig deep to find stuff on the internet that is now well known and ask a model about it,and for sure he knows a lot about it.
Put this thing in SillyTavern and you would get a chatbot that can play any role you can imagine,NSFW or non NSFW,freaking amazing.
Great video BTW.
The word "so" was used 426 times in this video! 25 "so" per minute! So, its look like Chris very loves the word "so". Hahaha)))
Is the model fine tuned in anyway to use the function calling convention (tags + json), or is it solely based off the system prompt? If the later, I am mighty impressed for a 8B model.
Full tutorial but code is pay walled, reminds me of "open"ai
That's amazing! What is the difference between using LangChain and not using it? Thanks!
This is amazing, thanks for sharing.
I love the content. I have an idea for a video and this is something a lot of people could use. I’m terrible at taking meeting notes. A local model that could listen using the laptop mic and summarize in person meetings would be amazing.
one of the best videos ive seen on this topic maybe the best! keep it up broski! <3
Mr. Ai, please create prompt installation