README.md 5.92 KB
Newer Older
Zhuohan Li's avatar
Zhuohan Li committed
1
2
<p align="center">
  <picture>
Zhuohan Li's avatar
Zhuohan Li committed
3
4
    <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/vllm-project/vllm/main/docs/source/assets/logos/vllm-logo-text-dark.png">
    <img alt="vLLM" src="https://raw.githubusercontent.com/vllm-project/vllm/main/docs/source/assets/logos/vllm-logo-text-light.png" width=55%>
Zhuohan Li's avatar
Zhuohan Li committed
5
6
  </picture>
</p>
Woosuk Kwon's avatar
Woosuk Kwon committed
7

Zhuohan Li's avatar
Zhuohan Li committed
8
9
10
<h3 align="center">
Easy, fast, and cheap LLM serving for everyone
</h3>
Woosuk Kwon's avatar
Woosuk Kwon committed
11

Zhuohan Li's avatar
Zhuohan Li committed
12
<p align="center">
13
| <a href="https://docs.vllm.ai"><b>Documentation</b></a> | <a href="https://vllm.ai"><b>Blog</b></a> | <a href="https://arxiv.org/abs/2309.06180"><b>Paper</b></a> | <a href="https://discord.gg/jz7wjKhh6g"><b>Discord</b></a> |
Woosuk Kwon's avatar
Woosuk Kwon committed
14

Zhuohan Li's avatar
Zhuohan Li committed
15
</p>
Woosuk Kwon's avatar
Woosuk Kwon committed
16

17
18
---

19
20
21
22
23
24
25
26
27
**The Fifth vLLM Bay Area Meetup (July 24th 5pm-8pm PT)**

We are excited to announce our fifth vLLM Meetup!
Join us to hear the vLLM's recent updates and the upcoming roadmap.
Additionally, our collaborators from AWS will be presenting their insights and experiences in deploying vLLM.
Register now [here](https://lu.ma/lp0gyjqr) and be part of the event!

---

Zhuohan Li's avatar
Zhuohan Li committed
28
*Latest News* 🔥
29
- [2024/06] We hosted [the fourth vLLM meetup](https://lu.ma/agivllm) with Cloudflare and BentoML! Please find the meetup slides [here](https://docs.google.com/presentation/d/1iJ8o7V2bQEi0BFEljLTwc5G1S10_Rhv3beed5oB0NJ4/edit?usp=sharing).
30
- [2024/04] We hosted [the third vLLM meetup](https://robloxandvllmmeetup2024.splashthat.com/) with Roblox! Please find the meetup slides [here](https://docs.google.com/presentation/d/1A--47JAK4BJ39t954HyTkvtfwn0fkqtsL8NGFuslReM/edit?usp=sharing).
31
32
- [2024/01] We hosted [the second vLLM meetup](https://lu.ma/ygxbpzhl) with IBM! Please find the meetup slides [here](https://docs.google.com/presentation/d/12mI2sKABnUw5RBWXDYY-HtHth4iMSNcEoQ10jDQbxgA/edit?usp=sharing).
- [2023/10] We hosted [the first vLLM meetup](https://lu.ma/first-vllm-meetup) with a16z! Please find the meetup slides [here](https://docs.google.com/presentation/d/1QL-XPFXiFpDBh86DbEegFXBXFXjix4v032GhShbKf3s/edit?usp=sharing).
Zhuohan Li's avatar
Zhuohan Li committed
33
- [2023/08] We would like to express our sincere gratitude to [Andreessen Horowitz](https://a16z.com/2023/08/30/supporting-the-open-source-ai-community/) (a16z) for providing a generous grant to support the open-source development and research of vLLM.
Lianmin Zheng's avatar
Lianmin Zheng committed
34
- [2023/06] We officially released vLLM! FastChat-vLLM integration has powered [LMSYS Vicuna and Chatbot Arena](https://chat.lmsys.org) since mid-April. Check out our [blog post](https://vllm.ai).
Zhuohan Li's avatar
Zhuohan Li committed
35
36

---
37
## About
Woosuk Kwon's avatar
Woosuk Kwon committed
38
vLLM is a fast and easy-to-use library for LLM inference and serving.
39

Zhuohan Li's avatar
Zhuohan Li committed
40
vLLM is fast with:
41

Zhuohan Li's avatar
Zhuohan Li committed
42
- State-of-the-art serving throughput
43
- Efficient management of attention key and value memory with **PagedAttention**
44
- Continuous batching of incoming requests
45
- Fast model execution with CUDA/HIP graph
Zhuohan Li's avatar
Zhuohan Li committed
46
- Quantization: [GPTQ](https://arxiv.org/abs/2210.17323), [AWQ](https://arxiv.org/abs/2306.00978), [SqueezeLLM](https://arxiv.org/abs/2306.07629), FP8 KV Cache
47
- Optimized CUDA kernels
Zhuohan Li's avatar
Zhuohan Li committed
48

49
50
**Performance benchmark**: We include a [performance benchmark](https://buildkite.com/vllm/performance-benchmark/builds/3924) that compares the performance of vllm against other LLM serving engines ([TensorRT-LLM](https://github.com/NVIDIA/TensorRT-LLM), [text-generation-inference](https://github.com/huggingface/text-generation-inference) and [lmdeploy](https://github.com/InternLM/lmdeploy)).

Zhuohan Li's avatar
Zhuohan Li committed
51
52
vLLM is flexible and easy to use with:

53
- Seamless integration with popular Hugging Face models
Zhuohan Li's avatar
Zhuohan Li committed
54
- High-throughput serving with various decoding algorithms, including *parallel sampling*, *beam search*, and more
55
- Tensor parallelism and pipeline parallelism support for distributed inference
56
57
- Streaming outputs
- OpenAI-compatible API server
58
- Support NVIDIA GPUs, AMD CPUs and GPUs, Intel CPUs and GPUs, PowerPC CPUs
Zhuohan Li's avatar
Zhuohan Li committed
59
60
- (Experimental) Prefix caching support
- (Experimental) Multi-lora support
61

62
63
64
65
66
67
68
69
vLLM seamlessly supports most popular open-source models on HuggingFace, including:
- Transformer-like LLMs (e.g., Llama)
- Mixture-of-Expert LLMs (e.g., Mixtral)
- Multi-modal LLMs (e.g., LLaVA)

Find the full list of supported models [here](https://docs.vllm.ai/en/latest/models/supported_models.html).

## Getting Started
70

71
Install vLLM with pip or [from source](https://vllm.readthedocs.io/en/latest/getting_started/installation.html#build-from-source):
Zhuohan Li's avatar
Zhuohan Li committed
72
73
74
75
76

```bash
pip install vllm
```

77
Visit our [documentation](https://vllm.readthedocs.io/en/latest/) to learn more.
78
79
80
- [Installation](https://vllm.readthedocs.io/en/latest/getting_started/installation.html)
- [Quickstart](https://vllm.readthedocs.io/en/latest/getting_started/quickstart.html)
- [Supported Models](https://vllm.readthedocs.io/en/latest/models/supported_models.html)
Zhuohan Li's avatar
Zhuohan Li committed
81

82
## Contributing
83

84
85
We welcome and value any contributions and collaborations.
Please check out [CONTRIBUTING.md](./CONTRIBUTING.md) for how to get involved.
Woosuk Kwon's avatar
Woosuk Kwon committed
86

87
88
89
90
91
92
93
94
95
96
97
98
99
100
## Sponsors

vLLM is a community project. Our compute resources for development and testing are supported by the following organizations. Thank you for your support!

<!-- Note: Please sort them in alphabetical order. -->
<!-- Note: Please keep these consistent with docs/source/community/sponsors.md -->

- a16z
- AMD
- Anyscale
- AWS
- Crusoe Cloud
- Databricks
- DeepInfra
101
- Dropbox
102
- Google Cloud
103
104
105
106
107
- Lambda Lab
- NVIDIA
- Replicate
- Roblox
- RunPod
108
- Sequoia Capital
109
110
111
- Trainy
- UC Berkeley
- UC San Diego
112
- ZhenFund
113
114
115

We also have an official fundraising venue through [OpenCollective](https://opencollective.com/vllm). We plan to use the fund to support the development, maintenance, and adoption of vLLM.

Woosuk Kwon's avatar
Woosuk Kwon committed
116
117
118
119
120
## Citation

If you use vLLM for your research, please cite our [paper](https://arxiv.org/abs/2309.06180):
```bibtex
@inproceedings{kwon2023efficient,
121
  title={Efficient Memory Management for Large Language Model Serving with PagedAttention},
Woosuk Kwon's avatar
Woosuk Kwon committed
122
123
124
125
126
  author={Woosuk Kwon and Zhuohan Li and Siyuan Zhuang and Ying Sheng and Lianmin Zheng and Cody Hao Yu and Joseph E. Gonzalez and Hao Zhang and Ion Stoica},
  booktitle={Proceedings of the ACM SIGOPS 29th Symposium on Operating Systems Principles},
  year={2023}
}
```