วีดีโอ

...for example... :)

how does it do in ARC and frontier math?

Drinking game ideas: every time he says ollama :D

is this api key is free or cost?

It can work on google collab ?

95% is zero, because each component will be 95%, therefore the error stacks up quickly to be unusable. Imagine a programming language that only gave 95% accuracy. It would be unusable.

is it possible to connect it with dify?

How to find tune with .md files as dataset? Most of the software documentation comes as markdown files. How to use them to fine tune models?

My question is how to safe the f16 merged directly as the safetensors, because it is getting binary .bin as native format ?

Can you alter the ai to give it a name

Any option to run this on Ollama, that is literally the cheapest! Thanks again for the tutorial! When you referencing the "original" directory, should all those directories referenced need to be in root and or in the original directory? The yaml is saying root.

Congratulations brother, hoping to see you in the first place.

btw ottodev might've performed better

i finally got to watch the whole vid. perhaps it'd have faired better with the parts that didn't copy? maybe that led to some confusion. also, a bit strange how the emoji confetti physics looked different between the qwen artifacts one and the "local" one... any idea why? maybe a temperature issue?

I hope they can and will implement this in Ollama ASAP. :-)

great examination of this, I was wanting to see how this worked. So llama is not only the real open ai, but also are seeminly actively trying to make it easy for people to use and modify it. I should probably look in to llama more.

Hi, Thanks for the video, btw I am facing issue with the last part where I could not able to access Salad endpoint and getting 403 Forbidden Error. Please let me know if you have anything for the same.

If i run the fastapi file using uvicorn then it will run on my local host or the machince local host.

But given that OpenAI, Claude, Llama, or any LLM can do this already, why do we need ollama for it?

Brilliant video. Examples with simple explanations thank you

Dang, what a great video

How can I give output of function back to LLM so, my final answer is from LLM instead of what function's returns.

This is the video for specific RP dataset creation that I have been needing. Please do a deep dive. Thanks.

Did not work. Errors. If execute Python ingest.

If I prompt "What is sky color ?" It answer using "Calling function: subtract_two_numbers" . It seems that llama3.2 is forced to use tool also if use tool has not meaning ?

I want to see a demo where the llm realises it doesn't have a tool it needs, clones a tools github project, adds the tool, commits it, and then uses it.

If I enter “hey, how's it going?” as a prompt, what result do I get?

Thanks for the clip and explanation. I see the fuction sometime takes number sometime it takes string, 1. how about concade "Prompt" + "Engineer" = "Prompt Engineer" , 2. Add Five + Two = 7. the place you wrote changing the char to int, kind of odd

❤awesome thanks for sharing

Things are advancing so fast... Now I've started using Windsurf, and wow, it feels like I'm a pro already. It's kind of scary because I don’t want to become overly reliant on it since I'm still new. I stumbled upon your channel almost (or over) a year ago, where you taught me the basics of chatbots. Thanks, man-without you, I wouldn't be where I am today.

investing in 3090 24gb a few years back was a good choice :D Local 32B baby

To be real, these small models are faster and better for many use cases.

hey there! i actually want to make a chatbot for my college project, I want to fine tune the LLM with my own dataset and then deploy the project with an UI but due to low computational power I am facing a lot problems in the deployment stage, could you possibly help me here? big thanks

thank you, very helpful.How to deploy NVIDIA'S AI models as API Using Flowise AI.

Thank you very much! The only guide I was able to use to create a bot You are the best!!!

Thank you for sharing the tutorial! I found it a bit cumbersome that Ollama requires additional installations for a graphical interface. However, the Page Assist you introduced seems much simpler since it only requires installing a Chrome extension to use. That said, when using a locally downloaded Ollama model through the browser, does it still count as running the LLM locally? Can it operate offline to avoid privacy exposure? Thank you!

I would like to suggest you, make your voice a little bit solfter use any ai tool, it will definitely help increase views and subscribers.

Thanks

This was nice

Hi I am connecting from another computer. I have my Ollama in AWS cloud. How do I make it where I can train it like what you did here?

forget conda and just use uv, much lighter and faster and installs in seconds via pip install uv

Funny accent 🤣😂 Speak english please

Does it work with local models?

Can 32B model run on RTX3090 config? It's really cheaper

That’s awesome! Thanks mate

in your next video, can you show how to connect this to a RAG, a Customer DB, Appointment setting, etc. Basically something more than just talking to the Ai.

Can we expect greater latency given multimodal?

Did you get your rag wireframe set up for big company use?

i'd recommend using uv instead of conda: pip install uv - uv pip install ... - reason: uv resolves module conflicts and has a bunch of other benefits, plus runs async. it can do venvs, init and manages a project's packages, ... it's my #1 python modules management tool replaced conda and pip itself. runs in jupyter also !pip install uv !uv pip install ...

hi there. even without installing the artifacts (v2) it shows the execution part...maybe the devs already implemented it inside the core.. ?