LLama is undoubtably one of the best open-source LLM out there by Meta. If you have ever wandered how it works, how these LLMs are able to produce human like text and talk like humans. Well, you are not alone. I am currently working towards my Masters thesis and learning the intricate of LLMs while documenting my learning journey.

These LLMs are made of something called "neural network," a term that may sound fancy, but essentially refers to a multitude of interconnected components collaborating harmoniously. These components work together to make decisions, enabling the model to perceive information and construct sentences based on the words they see.

What are Weights and Parameters?

Now, the real question is, what are weights and parameters. Well, as a visual learner myself i will take a pictographic approach to explain this two concepts in the simplest way possible.

Okay... Imagine you have a robot friend, and you want to teach this robot friend of yours how to recognize different animals. Now, lets say your robot friend has a backpack. In this backpack, there are some special tools called "weights" and "parameters."

Weights are like the knowledge or importance your robot friend gives to different features of an animal. You can think of weights like the power of connections between different parts of the robot friend's brain. For example: If you're teaching it to recognize a cat, the weight for features like pointy ears and a tail would be high because those are important for identifying a cat.

Parameters, on the other hand, parameters are like the settings your robot friend uses to make decisions. Imagine this like tuning a radio to get the best signal. Parameters help the your model adjust its thinking based on what it learns.

So, in simple terms, weights are how much your model cares about different things, and parameters are the settings it uses to make smart choices. Together, these help your model learn and make decisions, just like you do when you learn about new things.