Publicly open meta of big language model AI
Meta Platforms says they will open public access to their large language model for research on WHO (artificial intelligence).
Share on the blog, Metathe parent company of Facebookhope that the open release of the Open Pretrained Transformer (OPT-175B) model – a language model with up to 175 billion parameters – will help researchers understand and make good use of how language models work. big language.
 |
Meta opens access to large language models for AI research shutterstock |
A “big language model” is a collection of many natural language processing systems, trained on large volumes of text with the ability to answer questions or generate new text in response (in a reading comprehension rather than simply answering a fixed number of programming sentences).
Meta argued that restricting access to such models “hindered progress in efforts to improve their robustness…”, so they released the model under a non-commercial license to focus on research use cases, while preventing abuse and “maintaining integrity” for the model.
Audiences with access to the model, according to Meta, will include academic researchers affiliated with government, civil society or academic institutions; and industry research laboratories.
Besides the pre-trained models, Meta will provide the source code so that researchers and programmers can use them more easily.
at Blogtuan.info – Source: thanhnien.vn – Read the original article here
