Generative AI

My Gen AI Project

I am pretty sure all if you have used AI, examples being ChatGPT and Gemini; Or, Image generation AI's like Dall-E 3 and Meta's AI model(LLama3). My goal with this project is to run an AI model locally, on my dekstop, without utilizing the internet. To achieve this, are you realise a program called Ollama. With this I am able to run many models are available online, ranging from Google's Gemma model to Facebook's llama-3 model. There are also many other open source models are available on a website called hugging face. Similar to Github, hugging face is an online library for AI models. On this website, you can get models from companies such as Meta, Google, etc or you can download open source models that have other features such as image detection and custom text generation. The first model I chose for this project is Llama-2. My desktop houses a GTX-1660 Super, one which I use to work with AI. AI models come with something known as quantisation. This refers to the amount of connections and model House. The more the better. In this case, my desktop is not that powerful hence I chose the 7 billion version quantisation. In professional service, it is usually between 48 to 200,000,000,000. Recently Lama had been rereleased as 3.1 and offered a 786 billion quantisation version.