Using Ollama To Build a FULLY LOCAL 'ChatGPT Clone'

preview_player
Показать описание
In this video, I show you how to use Ollama to build an entirely local, open-source version of ChatGPT from scratch. Plus, you can run many models simultaneously using Ollama, which opens up a world of possibilities.

Enjoy :)

Join My Newsletter for Regular AI Updates 👇🏼

Need AI Consulting? ✅

Rent a GPU (MassedCompute) 🚀
USE CODE "MatthewBerman" for 50% discount

My Links 🔗

Media/Sponsorship Inquiries 📈

Links:
Рекомендации по теме
Комментарии
Автор

I'm building my own personal AI assistant but every time I start something a week later something better drops. My god, this is impossible. I've got to think better about my abstractions to make some of this stuff more drop-in ready. That might be an interesting video (or series of videos) for you Matthew, if not likely a bit advanced for your audience.

avi
Автор

😮 Please create a video integrating Ollama with autogen!

elierh
Автор

yes! it would be really interesting how autogen + Ollama goes !😍

taeyangoh
Автор

Every time I need something, you present a tool doing exactly that. Thanks!

rakly
Автор

Ollama series! This was a great starting video❤ thank you for all your hard work

xdasdaasdasd
Автор

I really like your content, but clicklying is the worst worm of clickbait. Please don't call it "installing ChatGPT" if you actually mean the generic version "Large Language Model". The title makes it seem like you are actually installing ChatGPT or a reverse engineered version of it, which implies the higher quality that ChatGPT has over standard open source models. You could call it "installing open source ChatGPT equivalent" or similar.

GoldenSunDS
Автор

Thank you so much Mathew, this is so incredible!

aldoyh
Автор

Tus vídeos y explicaciones son muy buenas.
Hay la posibilidad de liberar el potencial y poder entrenarlo para que pueda ser un modelo muy intelectual y articulador. Sin ninguna atadura ideologíca ni sesgo cognitivo por tenes que adecuarde a los canones políticamente correcto. Ina ia libre. Se puede?

wistonbritan
Автор

This is awesome! I'd love to see more. I feel like this can become something pretty robust with enough time.

MakilHeru
Автор

Building an AutoGen application using Ollama would be wonderful ! Example: one of the agents is a coder, implemented by a LLM specialized in coding etc.

scitechtalktv
Автор

Would have been nice to discuss what sort of resource demands these models put on your machine

magnusbrzenk
Автор

About the privateGPT, I found the accuracy can be improved if the database change from duckDB to elasticsearch.

fenix
Автор

This is just great and easy as well! Could you show us how to train these models with PDFs and Websites?

the.flatlander
Автор

Nice video! You definitely picked up a new subscriber here. I’d be interested in seeing how to build out a RAG solution with Ollama, and also how to make it run in parallel for multiple concurrent requests.

AlGordon
Автор

Can we combine this with fine-tuning where we first add a number of our own documents and then ask questions? NB I'd like to add the documents just once so that between sessions I can ask the model about these documents.

padonker
Автор

This was so Dope! - I have been using Ollama for a while, testing multiple models, and because of my lack of coding expertise, I had no understanding that it could be coded this way. I would like to see if you can use Ollama, memGPT, and Autogen, all working together 100% locally to choose the best model for a problem or question, call the model and get the result, and then permanently remember what is important about the conversation... I Double Dare You. ;)

LerrodSmalls
Автор

Awesome, could you integrate memgpt now :)

petersvideofile
Автор

Thank you for the info. Kindly, let us know what are the specs on your pc? I have a very slow response on my macbook air from 8GB Memory and CPU of M1

abdulazizalmass
Автор

this model will allow us to make open source models fast, I love the simultaneous part, please make more tutorials on this once it hits windows without wsl

WaefreBeorn
Автор

Great video, just a quick note, you actually do not need to all the previous messages and responses as the prompt, the API response contains an array of numbers called the context, just send that in the data of the next request

GutenTagLP