Don’t worry, there is a much clearer diagram at the end of the article.

You want to run LLMs(LLama, Mistral-7B) locally, and you have read tons of videos and articles on the Internet, but you are still in the maze.

Then, this article is the final one you need to get out of the maze.

I will sort out all the key concepts for you to understand them thoroughly, with ONE big diagram.

What is a Model in the context of LLM?

It’s nothing but a digital file, just like your image files and text files.

This is what a model exactly is:

It stores the structure and parameters of a neural network.

You know, all the modern AI technologies are based on neural networks. A neural network contains tons of nodes and each node has some parameters.

All of these parameters were stored in model files.

Also, you should notice:

• It stores parameters. But it does NOT contain training data.
• It’s NOT an executable program.

All we need to do in the future is:

• Download model files.
• Import them to our code.

How are these models trained?

To train a model, we need a large amount of data, to choose an algorithm, many GPUs, and enormous electricity.