It's actually a collection of models of varying sizes and training strategies. To set the scene for the rest of the course, I want to give you more details about the various models and how they were trained. Let's dive in. However, larger models are also more computationally expensive to train and deploy than the smaller models. Each of these models can be used for different application scenarios and purposes. The instruction-tuned models are created by taking the base models, also called foundation models, and running them through additional training called instruction tuning. This enables instruction-tuned models to better follow human language instructions such as summarize this or tell me a joke. These three instruction tune LLAMA models are called LLAMA chat models. Depending on your use case, you can take any of these models and further fine-tune them for your application needs. Although it's been more common to use the base foundation models for fine-tuning. With all the large language models out there, you may be wondering, how does LLAMA fit in? It's also worth noting how you are able to access these large language models. GPT 3.5 is accessible through API calls to OpenAI, which may also be fine for many of your use cases. Another cool thing about LLAMA models is the ecosystem of open source libraries It's called Lama CPP. In August of 2023, we released yet another Lama, Code LLAMA. For each size, there is a base version and also an instruct version. The base Code LLAMA models are derived from the non-chat LLAMA models. They can respond to human instructions, such as, help me write some code to build a web page, or please debug the following code that I just wrote. The Code LLAMA chat models generate code too, Pretty much all the most popular languages, including Python, JavaScript, C++, Java, HTML, and more. And there's one more LLAMA, Code LLAMA Python, One to ensure that AI generated code is safe against cybersecurity attacks. And one that checks if LLM inputs and outputs are safe, honest, and harmless. The first is called CyberSec Eval. It's a set of tools and also a benchmark data set that many are using to check if their code completion tools are generating secure code that guards against viruses or cyber threats. The other is yet another LLAMA model called LLAMA Guard, Let's go on to the next lesson.