BLOOM: 176-Billion Parameter Multilingual Language Model Ready For Open Use
The long awaited BLOOM has finished training and is ready for use and download by the public. This new model is the largest multilingual languages processing model to date.
Created on July 12|Last edited on July 13
Comment
For over a year, researchers from across the globe have been collaborating with BigScience to develop BLOOM, the world's largest open multilingual language model.
After a 117-day-long final training run completed on July 6th, the model finished training and was prepared for a public release of the final checkpoint. Today, BigScience has released everything, including an interactive demo, freely accessible through Hugging Face.
What is BLOOM?
BLOOM is a 175-billion parameter model for language processing, able to generate text much like GPT-3 and OPT-175B. It was developed to be multilingual, being deliberately trained on datasets containing 46 natural languages and 13 programming languages.
Luckily for the public, BLOOM is open-access, meaning anyone is able to download and use BLOOM for themselves. Everything about BLOOM is openly available on the various pages within BigScience's Hugging Face page, from the training logs to models of various sizes to checkpoints.
On the demo page, you can input your own string of text to start with, and the model will write the rest (even turning it into a nicely formatted image for you to share). It takes a little wrestling to get good results in my experience, but if you provide a good enough prompt, it can produce some impressively coherent and interesting passages.
Though the model was never deliberately trained on Japanese, it can also handle Japanese text. I would test other languages if I understood them, but I'm sure the results are pretty good in the supported languages.
Find out more
Add a comment
Tags: ML News
Iterate on AI agents and models faster. Try Weights & Biases today.