Our new open source and open access 7- and 70-billion parameter large language moddel (LLM for medicine, built on the back of Llama 🦙.
Building LLMs is expensive 💨💸 💸 💸 , which prices out the voices of low-resource settings.
Making Meditron an open-source, open-access public good, will allow researchers to validate the tool in a range of challenging and under-represented healthcare settings for a fraction of the cost and without reinventing the wheel 🎡
Check out the press release 👉 https://actu.epfl.ch/news/epfl-s-new-large-language-model-for-medical-knowle/
📜 Paper: https://arxiv.org/abs/2311.16079
🔗️ Code: https://github.com/epfLLM/meditron
🔗️ Distributed trainer: https://github.com/epfLLM/Megatron-LLM
🤗 Hugging Face hub: https://huggingface.co/epfl-llm
We are excited to continue working with the innovative team at the International Committee of the Red Cross (ICRC) [Javier Elkin, Blaise Robert and Fabrice Lauper] and our dream team of clinical guideline experts [Alix Miauton, Rainer Tan, Kristina Keitel..and so many more] as we work towards representative open validation through open models!
💰 Field studies cost money. Please reach out if you can recommend funders/donors interested in helping us scale this initiative of LLMs4 humanitarian response.
--
Finally, I really can’t help but get sentimental 🥹 about the phenomenally talented army of 30+ students and colleagues with whom I had the honor of working on this massive project. When you are making something to contribute to a public good, the energy is contagious 🌬️🌈
Zeming Chen Antoine Bonnet Alexandre Sallinen Martin Jaggi Antoine Bosselut Vinitra Swamy Matteo Pagliardini Angelika Romanou Syrielle Montariol Francesco Salvi ...and many more!