AI Paper Atlas

Research map / Large language models / AI and society

Estimating the Carbon Footprint of BLOOM, a 176B Parameter Language Model

Alexandra Sasha Luccioni, Sylvain Viguier, Anne-Laure Ligozat

JMLR 2023 · Large language models › AI and society · approach: Benchmarking and reproducibility practice

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
include idle consumption and embodied emissions in the footprintBLOOM training · BLOOM-176B trainingCO2eq (tonnes): 24.69 -> 50.5Mixed
train on low-carbon grid (57 gCO2eq/kWh) instead of OPT's grid (231 gCO2eq/kWh)BLOOM · LLM training footprintCO2eq (tonnes): 70 -> 25 tonnesImproved
train on low-carbon grid instead of GPT-3's grid (429 gCO2eq/kWh)BLOOM · LLM training footprintCO2eq (tonnes): 502 -> 25 tonnesImproved
compare raw energy of BLOOM and OPTBLOOM · LLM training footprintenergy (MWh): 324 -> 433 MWhHurt
deploy model as real-time API with unpredictable requestsBLOOM API · BLOOM API inferenceper-GPU power (W): 78-171W vs 400W TDPHurt
zero incoming requests to API instanceBLOOM API · BLOOM API inferenceenergy per 10 minutes (kWh): approximately 0.28 kWhNo clear effect
add intermediate model experiments (104B, 13B, 1B) to final model trainingBLOOM · BigScience experimentsCO2eq (tonnes): 35.8 tonnes vs 24.69Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy