#214 · Primary category: Inference & Local Deploy

ComfyUI_NetDist

Run ComfyUI workflows on multiple local GPUs/networked machines.

Project last updated:05/22/24

GitHub Stars

505

Forks

50

Contributors

1

License

Apache-2.0

Why we included this project

For people running ComfyUI jobs that stall on one card, NetDist spreads a workflow across the GPUs you already own or across machines on your local network. Queue nodes dispatch part of the graph to a second ComfyUI instance, whether that's another GPU on the same machine or a separate PC reached over the network, and a fetch node brings the finished output back. A two-node setup covers simple dual-GPU rigs, while the advanced mode can send an entire workflow JSON and lets you set the final image to "any" instead of a fixed output, which is handy once you scale past two cards. It suits power users and small teams who already manage multiple ComfyUI instances and want higher throughput or to render on a dedicated box without stitching pipelines together by hand. Installation is just pip install requests and a clone into the custom nodes folder, so it's easy to try on hardware you already have.

Articles for this project

No articles for this project yet.

To suggest a topic or contribute an article, contact us.

Related projects in this category