抑制碳排放：机器翻译中的碳排放基准评测
计算与语言
2022-04-05 v4 人工智能
机器学习
摘要
近年来，NLP领域取得了明确的进展，随着语言模型性能提升带来的效用增加，其应用也在不断扩展。然而，这些模型需要大量的计算能力和数据进行训练，从而导致巨大的碳足迹。因此，研究碳效率并寻找替代方案以减少训练模型（尤其是大语言模型（LLM））的总体环境影响势在必行。在我们的工作中，我们评估了多种语言对下机器翻译模型的性能，以评估为这些语言对训练模型所需计算能力的差异，并考察这些模型的各个组件，以分析我们流程中可被优化以减少碳排放的方面。
引用
@article{arxiv.2109.12584,
title = {Curb Your Carbon Emissions: Benchmarking Carbon Emissions in Machine Translation},
author = {Mirza Yusuf and Praatibh Surana and Gauri Gupta and Krithika Ramesh},
journal= {arXiv preprint arXiv:2109.12584},
year = {2022}
}
备注
The authors find these results limited as there are no clear, identifiable trends that provide useful information. We need/intend to make the experiments more robust as currently, they are not. The authors do not have access to the required computational sources for this at the moment. We will revisit the optimization of machine translation models using a single language pair at a later point