---
license: mit
---

Please refer to the [SepLLM paper - ICML 2025](https://arxiv.org/abs/2412.12094), [StreamingLLM paper - ICLR 2024](https://arxiv.org/abs/2309.17453), and our [`GitHub repository`](https://github.com/HKUDS/SepLLM)  for using this model.

To use the checkpoint of this model, you must install the `transformers-4.38.0.post1+sepllm-py3-none-any.whl` released from our [`GitHub repository`](https://github.com/HKUDS/SepLLM). Below are the reference script for testing and a sample of test results. We conducted testing using `lm_eval==0.4.0`.

```
CUDA_LAUNCH_BLOCKING=1
lm_eval --model hf \
	--model_args pretrained=Gausson/pythia-160m-deduped-n64-StreamingLLM \
	--tasks  arc_challenge,arc_easy,lambada_openai,logiqa,piqa,sciq,winogrande,wsc,wikitext  \
	--num_fewshot 5 \
	--device cuda:0\
	--batch_size 32
```

```
hf (pretrained=Gausson/pythia-160m-deduped-n64-StreamingLLM), gen_kwargs: (None), limit: None, num_fewshot: 5, batch_size: 32
|    Tasks     |Version|Filter|n-shot|    Metric     |   | Value |   |Stderr|
|--------------|------:|------|-----:|---------------|---|------:|---|------|
|arc_challenge |      1|none  |     5|acc            |↑  | 0.2056|±  |0.0118|
|              |       |none  |     5|acc_norm       |↑  | 0.2517|±  |0.0127|
|arc_easy      |      1|none  |     5|acc            |↑  | 0.4739|±  |0.0102|
|              |       |none  |     5|acc_norm       |↑  | 0.4478|±  |0.0102|
|lambada_openai|      1|none  |     5|acc            |↑  | 0.2672|±  |0.0062|
|              |       |none  |     5|perplexity     |↓  |44.0211|±  |1.5224|
|logiqa        |      1|none  |     5|acc            |↑  | 0.2212|±  |0.0163|
|              |       |none  |     5|acc_norm       |↑  | 0.2488|±  |0.0170|
|piqa          |      1|none  |     5|acc            |↑  | 0.6376|±  |0.0112|
|              |       |none  |     5|acc_norm       |↑  | 0.6349|±  |0.0112|
|sciq          |      1|none  |     5|acc            |↑  | 0.7570|±  |0.0136|
|              |       |none  |     5|acc_norm       |↑  | 0.7100|±  |0.0144|
|wikitext      |      2|none  |     5|bits_per_byte  |↓  | 0.9686|±  |   N/A|
|              |       |none  |     5|byte_perplexity|↓  | 1.9569|±  |   N/A|
|              |       |none  |     5|word_perplexity|↓  |36.2348|±  |   N/A|
|winogrande    |      1|none  |     5|acc            |↑  | 0.5335|±  |0.0140|
|wsc           |      1|none  |     5|acc            |↑  | 0.4327|±  |0.0488|
```

If you find our work helpful, please consider giving us a star ⭐ @ our  [`GitHub repository`](https://github.com/HKUDS/SepLLM) and citing our paper. We greatly appreciate your support 😄
```
@inproceedings{chen2025sepllm,
  title={{SepLLM: Accelerate Large Language Models by Compressing One Segment into One Separator}},
  author={Chen, Guoxuan and Shi, Han and Li, Jiawei and Gao, Yihang and Ren, Xiaozhe and Chen, Yimeng and Jiang, Xin and Li, Zhenguo and Liu, Weiyang and Huang, Chao},
  booktitle={International Conference on Machine Learning},
  year={2025},
  note={Also available at arXiv:2412.12094}
}
```