Simple and Clear: MINIMAL code and architecture changes. Only one up-and-down projection layer introduced, NO recurrent memory caching or sparse attention required.
Train Short, Test Long: NO performance drop on the sequences 4x~8x longer than the training ones (see here).
Continuous Length Extrapolation: Explicitly modeling the continuous dynamics of context window size during length extrapolation.

More details about long-text modeling with our CLEX can be found at the git repo.

Model Zoo

Model Name	Model Type	Starting Point	Train Data	Train Length	MAX Test Length
CLEX-7B-4K	base	LLaMA-2-7B	Redpajama-Book	4K	16K
CLEX-7B-Chat-4K	chat	CLEX-7B-4K	UltraChat	4K	16K
CLEX-7B-16K	base	LLaMA-2-7B	Redpajama-Book	16K	64K
CLEX-7B-Chat-16K (this checkpoint)	chat	CLEX-7B-16K	UltraChat	16K	64K

How to Use

bash

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("DAMO-NLP-SG/CLEX-7B-Chat-16K", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("DAMO-NLP-SG/CLEX-7B-Chat-16K", torch_dtype=torch.bfloat16, trust_remote_code=True)
inputs = tokenizer("What is CLEX?", return_tensors="pt")
sample = model.generate(**inputs, max_length=128)
print(tokenizer.decode(sample[0]))

Citation

If you find our project useful, hope you can star our repo and cite our paper as follows:

@article{damonlpsg2023clex,
  author = {Chen, Guanzheng and Li, Xin and Meng, Zaiqiao and Liang, Shangsong and Bing, Lidong},
  title = {CLEX: Continuous Length Extrapolation for Large Language Models},
  year = 2023,
  journal = {arXiv preprint arXiv:2310.16450},
  url = {https://arxiv.org/abs/2310.16450}
}