June 14, 2024
We present LayerSkip, an end-to-end solution to speed-up inference of large language models (LLMs). First, during training we apply layer dropout, with low dropout rates for earlier layers and higher dropout rates for later layers, and an early exit loss where all transformer layers share the same exit. Second, during inference, we show that this training recipe increases the accuracy of early exit at earlier layers, without adding any auxiliary layers or modules to the model. Third, we present a novel self-speculative decoding solution where we exit at early layers and verify and correct with remaining layers of the model. Our proposed self-speculative decoding approach has less memory footprint than other speculative decoding approaches and benefits from shared compute and activations of the draft and verification stages. We run experiments on different Llama model sizes on different types of training: pretraining from scratch, continual pretraining, finetuning on specific data domain, and finetuning on specific task. We implement our inference solution and show speedups of up to 2.16x on summarization for CNN/DM documents, 1.82x on coding, and 2.0x on TOPv2 semantic parsing task. We open source our code at https://github.com/facebookresearch/LayerSkip.
Written by
Akshat Shrivastava
Diana Liskovich
Basil Hosmer
Bram Wasti
Liangzhen Lai
Nas Mahmoud
Saurabh Agarwal
Ahmed Roman
Ahmed Aly
Beidi Chen
Publisher
ACL
June 27, 2024
Chris Cummins, Volker Seeker, Dejan Grubisic, Baptiste Rozière, Jonas Gehring, Gabriel Synnaeve, Hugh Leather
June 27, 2024
June 25, 2024
Elena Voita, Javier Ferrando Monsonis, Christoforos Nalmpantis
June 25, 2024
June 25, 2024
Min-Jae Hwang, Ilia Kulikov, Benjamin Peloquin, Hongyu Gong, Peng-Jen Chen, Ann Lee
June 25, 2024
June 14, 2024
Sheng-Chieh Lin, Akari Asai, Minghan Li, Barlas Oguz, Jimmy Lin, Scott Yih, Xilun Chen
June 14, 2024
Product experiences
Foundational models
Product experiences
Latest news
Foundational models