Stability AI has released a new text generation AI model group called StableLM Suite. This suite is designed to compete with other systems, such as OpenAI GPT-4 and is available as open-source on GitHub.
StableLM is trained on a large-scale version of the open-source dataset, The Pile, which incorporates information from various sources such as Wikipedia, Stack Exchange, and PubMed. Despite its small size, the model is effective on conversational and coding tasks, similar to OpenAI ChatGPT, after being trained on experimental datasets.
The StableLM-Alpha model is trained on a new dataset based on The Pile containing 1.5 trillion tokens, making it one of the largest models available in the industry.
Stability AI is committed to transparency and accessibility in AI design, and the StableLM suite is no exception. Developers are encouraged to freely inspect, use and adapt the StableLM base model for commercial or research purposes under the terms of the CC BY-SA- 4.0 license. This license requires attribution to Stability AI, providing a link to the license and indicating if any changes have been made.
In addition to the StableLM Suite, Stability AI has also released a set of fine-tuned research models using a combination of five recent open-source datasets for conversational agents. These models are for research use only and are released under a non-commercial CC BY-NC-SA 4.0 license under Stanford University’s Alpaca.
As the number of open-source text-generating models grows, Stability AI believes everyone should have a say in their design. By providing fine-grained access to the model, the company hopes to facilitate the development of interpretability and safety techniques beyond what is possible with closed models.
Stability AI plans to release more models in the near future, with models containing 15-65 billion parameters to be released soon. The company is excited to work with developers and researchers to roll out the StableLM Suite. It aligns with community efforts such as Open Assistant, an initiative to kick off the crowdsourced RLHF program and create open-source datasets for AI assistants.
Stability AI’s release of StableLM Suite significantly contributes to the open-source text-generating model community. With its commitment to transparency and accessibility, the company is paving the way for the development of safer and more reliable AI tools.