Team-provided Docker/Podman image for LingBot-World

[ChenCheng-976]

Hi,

Thanks for your nice work!

Could we provide an official Docker/Podman setup (Dockerfile or prebuilt image) for LingBot-World? Installation is currently brittle because Python/CUDA/PyTorch/flash-attn versions must match and the flash-attn build order matters. A pinned container + example run command would make it reproducible on GPU servers.

Best,

[dr-data]

Agree. Waiting for the official docker with parameters for different hardware/GPU

[huzhanxiaoxue]

Agree，2

[art-from-the-machine]

I'm not affiliated with the team, but I have set up my own Dockerfile to run the quantized versions of these models behind a small Flask API:

https://huggingface.co/art-from-the-machine/lingbot-world-base-cam-nf4-server/blob/main/Dockerfile

I also ran into a few issues getting the different installations to play nicely, so hopefully this can help save time for others looking to get started / may be useful as a starting point for an official container.