Troubleshooting

This page lists some common errors and tips for fixing them.

CUDA error: an illegal memory access was encountered

This error may be due to kernel errors or out-of-memory issues.

• If it is a kernel error, it is not easy to fix.

• If it is out-of-memory, sometimes it will report this error instead of “Out-of-memory.” In this case, try setting a smaller value for --mem-fraction-static. The default value of --mem-fraction-static is around 0.8 - 0.9. https://github.com/sgl-project/sglang/blob/1edd4e07d6ad52f4f63e7f6beaa5987c1e1cf621/python/sglang/srt/server_args.py#L92-L102

The server hangs

If the server hangs, try disabling some optimizations when launching the server.

• Add --disable-cuda-graph.

• Add --disable-flashinfer-sampling.