Is it possible to setup the node to act like "cuda visible device" in SWARMUI?

Is it possible to set up each GPU to load their own separate checkpoint and continue their own workflow?

Similar to when you open four instances of A1111 and set the CUDA visible device to 0-3 for each instance.

Having 4 GPUs working for multiple tasks in parallel on the same page is what I mean.

I don't familiar with Comfy and can't figure it out how to make three distributed node doing that.

That is already happening for the workers.

The cuda visible part is already happening for the worker but four GPU working in the same flow to produce different seed.

Is there any way to select a specific gpu to load a specific checkpoint and continue their own flow?

Example.

Gpu0 load SDXLA to generate imageA

Gpu1 load sd1.5 to do imagetoimage

Gpu2 load SDXLB to generate imageB

Gpu3 load sd1.5 to do imagetoimage

Navigation Menu