Files
sglang/docs/flashinfer.md
2024-03-11 05:49:27 -07:00

19 lines
511 B
Markdown

## Flashinfer Mode
[flashinfer](https://github.com/flashinfer-ai/flashinfer) is a kernel library for LLM serving.
It can be used in SGLang runtime to accelerate attention computation.
### Install flashinfer
See https://docs.flashinfer.ai/installation.html.
### Run a Server With Flashinfer Mode
Add `--enable-flashinfer` argument to enable flashinfer when launching a server.
Example:
```bash
python -m sglang.launch_server --model-path meta-llama/Llama-2-7b-chat-hf --port 30000 --enable-flashinfer
```