The EASIEST way to run MULTIMODAL AI Locally! (Ollama ❤️ LlaVA)

With the power of Llava Models and Thanks to Ollama's support, you can run GPT-4 Vision like (not the exact match) Mutlimodal models locally on your computers (does not need CPU).

🔗 Links 🔗

My previous Ollama Tutorial (Web UI)

❤️ If you want to support the channel ❤️
Support here:

🧭 Follow me on 🧭

[Arrowtake]

This is great, got it running on my Mac M2 pro 32GB and it is fast! Thanks for the insight how to get this running

[s-informationatyourservi]

😂i bet a whole dollar there’s gonna be a fine tuned spicy model named butLava just in time for christmas😂❤~ thank you for another awesome tutorial

[zyxwvutsrqponmlkh]

Thanks for the tutorial, don't have the time to try it tonight but maybe tomorrow. Never eaten llama meat, but I do like baklava.

[kachunchau]

hi, Dr. 1LC, I think your work will help me to do some classification task experiment. So I don't want use it like in a chatbot way. How can I call some functions to get result? Hope you can see this comment

[haraldwolte]

When I run "ollama run Llava" it responds in Chinese. How to make it english? *Edit; updating ollama fixed it.
To update ollama, just run the ollama install script again

[MichealScott]

GPT 4.5pro released

what model is this. precise name - what is it called in the API?

ChatGPT
This model is called "ChatGPT with browsing", and in the context of the OpenAI API, it is referred to as "gpt-4.5-turbo". This version includes enhanced capabilities like browsing the internet, which is not available in the standard GPT-4 model. The browsing feature allows the model to pull in information from the web in real-time to provide more current and detailed responses.

User
Can you continue this text?
GPT models: these are various versions of the Generative Pre-trained
Transformer models. They include:

Gpt-3.5-turbo: an optimized version of the GPT 3.5 model
ChatGPT
Certainly! Continuing from where you left off:

GPT models: these are various versions of the Generative Pre-trained Transformer models. They include:

GPT-3.5-turbo: An optimized version of the GPT-3.5 model, designed for faster response times and more efficient processing. It maintains the advanced language understanding and generation capabilities of GPT-3.5 but with enhanced performance.

GPT-4: This model represents a significant leap in the scale and capabilities over its predecessors. It features a larger number of parameters, which allows for more nuanced and complex language processing. GPT-4 is known for its improved understanding of context and ability to handle more intricate and detailed tasks.

GPT-4.5-turbo (ChatGPT with browsing): A more recent iteration, this model combines the advanced capabilities of GPT-4 with the added functionality of real-time web browsing. This enables the model to access and incorporate up-to-date information from the internet, enhancing its responses with current data and expanded knowledge.

GPT-NEO and GPT-J: These are models developed by the open-source community, inspired by the GPT architecture. GPT-NEO and GPT-J aim to provide similar capabilities to OpenAI's versions, focusing on accessibility and open-source collaboration. They have been widely adopted for various applications where open-source models are preferred.

Each version of the GPT series builds upon the learnings and advancements of its predecessors, aiming to provide more sophisticated, accurate, and contextually relevant language generation and understanding. These models find applications in a wide range of fields, including customer service, content creation, programming assistance, and more, showcasing the versatility and power of AI in natural language processing.

[timothywcrane]

I think you meant "does not need GPU"... if not, I want to know how.

[雪鷹魚英語培訓的領航]

The pixel art in your video thumbnail would be a better channel avatar icon IMO.

[MichealScott]

❤

btw i got m1 mac and base version 8gb ram and it lags soo much if apps are on or not :( atleast who would be trying i must recommend them 16gb ram else it would be janky as hell!

[lennard]

I found moondream to be better than the small version of llava.

[etherhealingvibes]

whats the point?. poor AI models in your computer for what????

[costgorunner]

Mistral is the way to go.. cutting edge