Lucie 7B LLM training environmental impact assessed

A new life cycle assessment (LCA) details the environmental impact of training the Lucie 7B open-source large language model on the Jean Zay supercomputer. The study, which includes manufacturing emissions, operational energy, water consumption, and hardware infrastructure, reports a total training footprint of 21 tCO2eq for Lucie 7B. The Jean Zay H100 partition has an annual footprint of 417.5 tCO2eq, with an effective intensity of 36.7 gCO2eq per GPU-hour. The research also highlights water consumption and waste-heat recovery efforts, contributing to the understanding of frugal AI systems. AI

IMPACT Provides a detailed environmental footprint for LLM training, informing the development of more sustainable AI infrastructure.

RANK_REASON Academic paper detailing a life cycle assessment of LLM training. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• IDRIS
• Jean Zay Supercomputer
• Lucie 7B
• NVIDIA H100
• OpenLLM-France

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →