September 27, 2023
We present a series of long-context LLMs that support effective context windows of up to 32,768 tokens. Our model series are built through continual pretraining from LLAMA 2 with longer training sequences and on a dataset where long texts are upsampled. We perform extensive evaluation on language modeling, synthetic context probing tasks, and a wide range of research benchmarks. On research benchmarks, our models achieve consistent improvements on most regular tasks and significant improvements on long-context tasks over LLAMA 2. Notably, with a cost-effective instruction tuning procedure that does not require human-annotated long instruction data, the 70B variant can already surpass gpt-3.5-turbo-16k’s overall performance on a suite of long-context tasks. Alongside these results, we provide an in-depth analysis on the individual components of our method. We delve into LLAMA’s position encodings and discuss its limitation in modeling long dependencies. We also examine the impact of various design choices in the pretraining process, including the data mix and the training curriculum of sequence lengths – our ablation experiments suggest that having abundant long texts in the pretrain dataset is not the key to achieving strong performance, and we empirically verify that long context continual pretraining is more efficient and similarly effective compared to pretraining from scratch with long sequences.
Written by
Wenhan Xiong
Igor Molybog
Hejia Zhang
Rui Hou
Louis Martin
Yashar Mehdad
Sharan Narang
Kshitiz Malik
Mike Lewis
Jingyu Liu
Publisher
Meta
Research Topics
October 04, 2024
Bandhav Veluri, Benjamin Peloquin, Bokai Yu, Hongyu Gong, Shyam Gollakota
October 04, 2024
October 03, 2024
David Dale, Marta R. Costa-jussa
October 03, 2024
September 26, 2024
Belen Alastruey, Gerard I. Gállego, Marta R. Costa-jussa
September 26, 2024
September 05, 2024
Chunting Zhou, Lili Yu, Arun Babu, Kushal Tirumala, Michihiro Yasunaga, Leonid Shamis, Jacob Kahn, Luke Zettlemoyer, Omer Levy, Xuezhe Ma
September 05, 2024
Foundational models
Latest news
Foundational models