Scaling Up Models and Data with t5x and seqio
Affiliation: Lead Authors Affiliation: Adam Roberts Affiliation: Hyung Won Chung Affiliation: Anselm Levskaya Affiliation: Gaurav Mishra Affiliation: James Bradbury Affiliation: Technical Contributors Affiliation: Daniel Andor, Sharan Narang, Brian Lester, Colin Gaffney, Afroz Mohiuddin, Affiliation: Curtis Hawthorne, Aitor Lewkowycz, Alex Salcianu, Marc van Zee, Jacob Austin, Sebastian Goodman, Livio Baldini Soares, Haitang Hu, Sasha Tsvyashchenko, Aakanksha Chowdhery, Jasmijn Bastings, Jannis Bulian, Xavier Garcia Affiliation: Jianmo Ni, Andrew Chen, Kathleen Kenealy, Jonathan H. Clark, Stephan Lee Affiliation: Dan Garrette, James Lee-Thorp Affiliation: Technical Advisors Affiliation: Colin Raffel, Noam Shazeer, Marvin Ritter, Maarten Bosma, Alexandre Passos, Jeremy Maitin-Shepard Affiliation: Leadership Affiliation: Noah Fiedel, Mark Omernick, Brennan Saeta, Ryan Sepassi, Affiliation: Alexander Spiridonov, Joshua Newlan, Andrea Gesmundo Affiliation: *Authors are ordered by impact within groups.
Abstract
Recent neural network-based language models have benefited greatly from scaling up the size of training datasets and the number of parameters in the models themselves. Scaling can be complicated due to various factors including the need to distribute computation on supercomputer clusters (e.g., TPUs), prevent bottlenecks when infeeding data, and ensure reproducible results. In this work, we present two software libraries that ease these issues: t5x simplifies the process of building and training large language models at scale while maintaining ease of use, and seqio provides a task-based API for simple creation of fast and reproducible training data and evaluation pipelines. These open-source libraries have been used to train models with hundreds of billions of parameters on datasets with multiple terabytes of training data. Along with the libraries, we release configurations and instructions for T5-like encoder-decoder models as well as GPT-like decoder-only architectures.
原文 arXiv:2203.17189;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2203.17189v1