StableLM is here! Open Source and Commercial Use (Quick Setup)

Introduction to Stable LM Suite of Language Models

In this section, the speaker introduces the Stable LM Suite of Language Models and discusses its features.

Stable LM Alpha Release

• Stability AI has released their own large language model called Stable LM.

• It is completely free and open source, including commercially viable.

• The model comes in a 3 billion parameter and a 7 billion parameter version, with larger models coming soon.

• Developers can freely inspect, use, and adapt the base models for commercial or research purposes subject to the terms of the ccby essay 4.0 license.

History of Stability AI

• In 2022, Stability AI released Stable Diffusion, an image model that represents a transparent open and scalable alternative to proprietary AI.

• They are continuing to make foundational AI technology accessible to all through their work on Stable LM.

Features of Stable LM

• Stable LM models can generate text and code and will power a range of downstream applications.

• The models are trained on an experimental dataset built on the pile which is three times larger than previous datasets used for training language models.

• Despite its small size (3 to 7 billion parameters), stable LM has surprisingly high performance in conversational encoding tasks.

Using Stable LM Base Model Alpha 7B

In this section, the speaker demonstrates how to use the stable LM base model alpha 7B by providing instructions on how to install Nvidia SMI, bits and bytes transformers, accelerate modules.

Installing Required Modules

• To power the models via GPU you need to install Nvidia SMI

• Install bits and bytes Transformers and Accelerate which allows us to take this model and actually run it on these GPUs.

Setting Up the Interface

• Set up what the interface is going to look like and import the different modules that we're going to use.

Choosing a Model

• Choose your model, for example, stable LM base Alpha 7B or tuned Alpha 7B.

• Download all of this which may take some time depending on the size of the model.

Google Collab Notebook

• A Google collab notebook is available with code and instructions on how to get started.

• The speaker demonstrates how to use it in this section.

Introduction to Artificial Intelligence

In this section, the speaker introduces artificial intelligence and attempts to write Python code to count to 100.

Writing Python Code

• The speaker writes Python code using the import numpy as MP command to count up to 100.

Tuned Version of AI

• The speaker downloads Stable LM tuned Alpha 7B and tries writing a poem about artificial intelligence but gets an unexpected output.

• The base model performs better than the tuned model when asked questions about artificial intelligence.

Conclusion

• The speaker invites viewers to play around with AI and share their results in the comments. They also encourage viewers to learn more about AI and new technologies by sharing the video with friends and family.

• Finally, they thank viewers for watching and ask them to like and subscribe for more content on AI.