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
Ahmed Aly
Beidi Chen
Ahmed Roman
Nas Mahmoud
Saurabh Agarwal
Publisher
ACL
November 20, 2024
Jay Shah, Ganesh Bikshandi, Ying Zhang, Vijay Thakkar, Pradeep Ramani, Tri Dao
November 20, 2024
November 20, 2024
Igor Fedorov, Kate Plawiak, Lemeng Wu, Tarek Elgamal, Naveen Suda, Eric Smith, Hongyuan Zhan, Jianfeng Chi, Yuriy Hulovatyy, Kimish Patel, Zechun Liu, Yangyang Shi, Tijmen Blankevoort, Mahesh Pasupuleti, Bilge Soran, Zacharie Delpierre Coudert, Rachad Alao, Raghuraman Krishnamoorthi, Vikas Chandra
November 20, 2024
November 19, 2024
Shehzaad Dhuliawala, Ilia Kulikov, Ping Yu, Asli Celikyilmaz, Jason Weston, Sainbayar Sukhbaatar, Jack Lanchantin
November 19, 2024
November 14, 2024
Zhaoyu Li, Jialiang Sun, Logan Murphy, Qidong Su, Zenan Li, Xian Zhang, Kaiyu Yang, Xujie Si
November 14, 2024
Foundational models
Latest news
Foundational models