Effective Long-Context Scaling of Foundation Models
- URL: http://arxiv.org/abs/2309.16039v3
- Date: Tue, 14 Nov 2023 01:40:13 GMT
- Title: Effective Long-Context Scaling of Foundation Models
- Authors: Wenhan Xiong, Jingyu Liu, Igor Molybog, Hejia Zhang, Prajjwal
Bhargava, Rui Hou, Louis Martin, Rashi Rungta, Karthik Abinav Sankararaman,
Barlas Oguz, Madian Khabsa, Han Fang, Yashar Mehdad, Sharan Narang, Kshitiz
Malik, Angela Fan, Shruti Bhosale, Sergey Edunov, Mike Lewis, Sinong Wang,
Hao Ma
- Abstract summary: We present a series of long-context LLMs that support effective context windows of up to 32,768 tokens.
Our models achieve consistent improvements on most regular tasks and significant improvements on long-context tasks over Llama 2.
- Score: 90.57254298730923
- License: http://creativecommons.org/licenses/by/4.0/
- Abstract: 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.
Related papers
- LongSkywork: A Training Recipe for Efficiently Extending Context Length in Large Language Models [61.12177317970258]
LongSkywork is a long-context Large Language Model capable of processing up to 200,000 tokens.
We develop two novel methods for creating synthetic data.
LongSkywork achieves outstanding performance on a variety of long-context benchmarks.
arXiv Detail & Related papers (2024-06-02T03:34:41Z) - Long Context is Not Long at All: A Prospector of Long-Dependency Data for Large Language Models [13.091271774417867]
Long-context modeling capabilities are important for large language models (LLMs) in various applications.
We propose a data mining framework textbfProLong that can assign each training sample with a long dependency score.
Comprehensive experiments on multiple benchmarks indicate that ProLong effectively identifies documents that carry long dependencies.
arXiv Detail & Related papers (2024-05-28T07:36:56Z) - Dataset Decomposition: Faster LLM Training with Variable Sequence Length Curriculum [30.46329559544246]
We introduce dataset decomposition, a novel variable sequence length training technique.
We train an 8k context-length 1B model at the same cost as a 2k context-length model trained with the baseline approach.
Experiments on a web-scale corpus demonstrate that our approach significantly enhances performance on standard language evaluations and long-context benchmarks.
arXiv Detail & Related papers (2024-05-21T22:26:01Z) - Long Context Alignment with Short Instructions and Synthesized Positions [56.1267385315404]
This paper introduces Step-Skipping Alignment (SkipAlign)
It is a new technique designed to enhance the long-context capabilities of Large Language Models (LLMs)
With a careful selection of the base model and alignment datasets, SkipAlign with only 6B parameters achieves it's best performance and comparable with strong baselines like GPT-3.5-Turbo-16K on LongBench.
arXiv Detail & Related papers (2024-05-07T01:56:22Z) - LongWanjuan: Towards Systematic Measurement for Long Text Quality [102.46517202896521]
LongWanjuan is a dataset specifically tailored to enhance the training of language models for long-text tasks with over 160B tokens.
In LongWanjuan, we categorize long texts into holistic, aggregated, and chaotic types, enabling a detailed analysis of long-text quality.
We devise a data mixture recipe that strategically balances different types of long texts within LongWanjuan, leading to significant improvements in model performance on long-text tasks.
arXiv Detail & Related papers (2024-02-21T07:27:18Z) - LongAlign: A Recipe for Long Context Alignment of Large Language Models [61.85923382850057]
LongAlign is a recipe of the instruction data, training, and evaluation for long context alignment.
We construct a long instruction-following dataset using Self-Instruct.
We adopt the packing and sorted strategies to speed up supervised fine-tuning on data with varied length distributions.
arXiv Detail & Related papers (2024-01-31T18:29:39Z) - E^2-LLM: Efficient and Extreme Length Extension of Large Language Models [74.1254067728251]
We propose an Efficient and Extreme length extension method for Large Language Models, called E 2 -LLM, with only one training procedure and dramatically reduced cost.
Comprehensive experimental results on multiple benchmark datasets demonstrate the effectiveness of our E 2 -LLM on challenging long-context tasks.
arXiv Detail & Related papers (2024-01-13T02:11:20Z) - Adapting Pretrained Text-to-Text Models for Long Text Sequences [39.62224414485055]
We adapt an existing pretrained text-to-text model for long-sequence inputs.
We build a long-context model that achieves competitive performance on long-text QA tasks.
arXiv Detail & Related papers (2022-09-21T00:41:07Z)
This list is automatically generated from the titles and abstracts of the papers in this site.
This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.