Currently the Load PuLID ✦ Flux.2 node does not allow device selection defaulting to whatever GPU is loaded. The problem is that I think PuLID needs to be on the same GPU as the Flux2 model. If the Flux2 model is spread out over devices or loaded on a secondary GPU the workflow errors. I understand that there is support for older PuLID load nodes, but the output does not connect to the Apply PuLID Flux.2 node. Likely because it was designed for SDXL or a different model. Thanks.
Currently the Load PuLID ✦ Flux.2 node does not allow device selection defaulting to whatever GPU is loaded. The problem is that I think PuLID needs to be on the same GPU as the Flux2 model. If the Flux2 model is spread out over devices or loaded on a secondary GPU the workflow errors. I understand that there is support for older PuLID load nodes, but the output does not connect to the Apply PuLID Flux.2 node. Likely because it was designed for SDXL or a different model. Thanks.