Effective Long-Context Scaling of Foundation Models
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 mo...
Gespeichert in:
Hauptverfasser: | , , , , , , , , , , , , , , , , , , , , |
---|---|
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | 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. |
---|---|
DOI: | 10.48550/arxiv.2309.16039 |