Run Your Own ChatGPT-like LLM on Your Windows PC!

Thanks to a new project call LM Studio, it is now possible to run your own ChatGPT-like AI chatbot on your Windows PC. The app provides an easy way to download models like Facebook's LLAMA 2, and start chatting with them immediately. No cloud, no fees, all local!
---

#garyexplains

[magicmanchloe]

Can I just say that thumbnail is amazing!!

[briannewman]

A video on code instruct would be very useful.

[rajagopalan]

Can you share us a thought on building large language models specific to an industry that can be run locally? Also a video on code generation locally will be great. Thanks.

[TheGrizz]

A bunch of future videos on LLM seem like a good idea. I see many good ideas in comments.
my contribution:
Benchmark (different CPUs) (memory speed)
Different GPUs
Running on Gaming laptops since they are super popular

[cvicracer]

Yes please do a video on code instruct.

[silasdenny]

Yes please tell us about code instruct.

[michielsmissaert]

Great video! If you can run a large LLM on your pc you can also run it on a hosting service, right? How would this work if you want to use the LLM for AI voice calling?

[richardo]

Hi it would be nice to see the code instruct case. Is it possible for these model to interact with with applications on your machine such as ms excel. A use case will be where you give it a task and it opens excel, writes to the file and save to a file location, or open an existing file, read some data and interpret the information. Please help.

[jeveremattthews]

I can't load my model because I keep getting an error; error load model exit code: 1

[Shmack_]

Will all the restrictions be removed or no?

[jokmong]

Hi Gary u ar the best, im newbie and have a question, i exceed the token, is there any way we can get more tokens or do we have to wait like how long? thanks Gary

[Dan-zdqr]

I see CPU-only locally run Windows AI chatbots gaining more and more acceptance because many people simply can't afford expensive GPUs. This will get accepted from nerds to more mainstream users in the same way PCs first arrived on the scene. I'm looking forward to more open-source LLMs and effective speed-up tweaks for CPU-only vids.

[AndersHass]

Not that signed necessarily means it is safe either but chances are slimmer.
I do hope they just open source it once they are ready for that just like LLaMA itself.

[aspectcarl]

Please tell us about code construct Gary 🙂

[GetRealwithMike]

Hi Gary, can you make an episode using this software to create multiple agents on our own machine that will create programs? The lead agent suggests a variety of projects based on keywords (ex: automation of business cards) and the other agents create a program that can do it. That would be fun to have on my pc.

[Hobbies-and-Experiments]

This is amazing, thanks for uploading this, I followed your instructions and omg local Ai 👍🍻

[RandomStuffPT]

I need this but with code, not a no code tool.
LLM localy with any language

[MatthewHarrold]

I just installed this on my MBA 13", wish I had more RAM, but about 17% downloaded at 12:33am. Might be a Saturday project, going to bed. $0.02

[GetRealwithMike]

I def want to learn the Code one. I'd like to make a Chat one to help people that are shutins - people afraid to go out. They would have something to become a virtual friend with. How would someone with little coding abilities teach a model?

Say I want to make something that builds a desktop application for Windows and MAC, make it executable by clicking an icon, how to I load a coding language into it? It would be great if I could teach it any subject using PDFs or use scholarly articles about psychology, for example for it to learn psychology. Then make it capable of going on the web to find information if information isn't in the model. Can you show how to do that? That would be so amazingly helpful and I'd be able to terminate my ChatGPT subscription.

[raduboboc]

I get the infamous error: Failed to load model <*> Error: Error loading model. Exit code: 1 for all models that I've tried
I have a very beefy laptop with a CPU: Intel(R) Core(TM) i9-10885H CPU ( yes it has AVX2) + 64Gb of ram + GPU NVIDIA Quadro T2000 with Max-Q Design
I have installed the c++ requirements, I have AVX2, I have a good enough CPU, the latest version of windows 10, the 0.25 version of LM studio, latest gpu drivers
I have tried smaller and bigger models, but nothing works and I am stuck with the above error :(