Stability AI has released a new free and open-source language model called StableLM.
The model is available in three different parameter sizes for the Alpha phase: three billion, seven billion, fifteen billion, and sixty-five billion. Under the CC BY-SA-4.0 license rules, developers can review, utilize, and modify StableLM basic models for personal and commercial projects.
The StableLM models are trained on a dataset that is three times larger than The Pile, a popular open-source dataset used to train other language models. This allows StableLM to achieve unexpectedly excellent performance on conversational and coding tasks, despite only having 3–7 billion parameters (GPT-3 has 175 billion).
StableLM is an open system that allows researchers to analyse its performance, evaluate its results, detect dangers, and put safety measures in place. The models are intended for daily usage on a variety of devices, allowing developers to create programmes that function on a broader range of hardware. Instead of replacing consumers, the idea is to support them.
In addition to the Alpha models, Stability AI also released a collection of research models optimized for use in classroom settings. These models are fine-tuned using data from five recently released open-source conversational agent datasets. They are available under a noncommercial CC BY-NC-SA 4.0 license for academic research.
The sources for this piece include an article in MarktechPost.
