README.md

---
license: apache-2.0
---
license: apache-2.0
---

**Paper**: [Adapting Language Models to Compress Contexts](https://arxiv.org/abs/2305.14788)  

**Code**: https://github.com/princeton-nlp/AutoCompressors

**Models**: 
- Llama-2-7b fine-tuned models: [AutoCompressor-Llama-2-7b-6k](https://huggingface.co/princeton-nlp/AutoCompressor-Llama-2-7b-6k/), [FullAttention-Llama-2-7b-6k](https://huggingface.co/princeton-nlp/FullAttention-Llama-2-7b-6k)
- OPT-2.7b fine-tuned models:  [AutoCompressor-2.7b-6k](https://huggingface.co/princeton-nlp/AutoCompressor-2.7b-6k), [AutoCompressor-2.7b-30k](https://huggingface.co/princeton-nlp/AutoCompressor-2.7b-30k), [RMT-2.7b-8k](https://huggingface.co/princeton-nlp/RMT-2.7b-8k), [FullAttention-2.7b-4k](https://huggingface.co/princeton-nlp/FullAttention-2.7b-4k)
- OPT-1.3b fine-tuned models: [AutoCompressor-1.3b-30k](https://huggingface.co/princeton-nlp/AutoCompressor-1.3b-30k), [RMT-1.3b-30k](https://huggingface.co/princeton-nlp/RMT-1.3b-30k)

---

AutoCompressor-2.7b-6k is a model fine-tuned from [facebook/opt-2.7b](https://huggingface.co/facebook/opt-2.7b) following the AutoCompressor method in [Adapting Language Models to Compress Contexts](https://arxiv.org/abs/2305.14788). 
This model is fine-tuned on 2B tokens from [The Pile](https://pile.eleuther.ai). The pre-trained OPT-2.7b model is fine-tuned on sequences of 6,144 tokens with 50 summary vectors, summary accumulation, randomized segmenting, and stop-gradients. 

To get started, download the [`AutoCompressor`](https://github.com/princeton-nlp/AutoCompressors) repository and load the model as follows:

```
from auto_compressor import AutoCompressorModel

model = AutoCompressorModel.from_pretrained("princeton-nlp/AutoCompressor-2.7b-6k")
``` 

**Evaluation**

We record the perplexity achieved by our OPT-2.7b models on segments of 2048 tokens, conditioned on different amounts of context. 
FullAttention-2.7-4k  uses full uncompressed contexts whereas AutoCompressor-2.7b-6k and RMT-2.7b-8k compress segments of 2048 tokens into 50 summary vectors.

*In-domain Evaluation*

| Context Tokens               | 0   |512  | 2048 | 4096 | 6144 |
| -----------------------------|-----|-----|------|------|------|
| FullAttention-2.7b-4k        | 6.57|6.15 |5.94  |-     |-     |
| RMT-2.7b-8k                  | 6.34|6.19 |6.02 | 6.02  | 6.01 |
| AutoCompressor-2.7b-6k       | 6.31|6.04 | 5.98 | 5.94 | 5.93 | 

*Out-of-domain Evaluation*

| Context Tokens               | 0   |512  | 2048 | 4096 | 6144 |
| -----------------------------|-----|-----|------|------|------|
| FullAttention-2.7b-4k        | 8.94|8.28 |7.93  |-     |-     |
| RMT-2.7b-8k                  | 8.62|8.44 |8.21 | 8.20  | 8.20 |
| AutoCompressor-2.7b-6k       | 8.60|8.26 | 8.17 | 8.12 | 8.10 | 

See [Adapting Language Models to Compress Contexts](https://arxiv.org/abs/2305.14788) for more evaluations, including evaluation on 11 in-context learning tasks.

## Bibtex
```
@misc{chevalier2023adapting,
      title={Adapting Language Models to Compress Contexts}, 
      author={Alexis Chevalier and Alexander Wettig and Anirudh Ajith and Danqi Chen},
      year={2023},
      eprint={2305.14788},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}
```