invoke-ai
[enhancement]: Flux controlnet models
Is there an existing issue for this?
- [X] I have searched the existing issues
Contact Details
No response
What should this feature add?
Just a question, how do you add flux control nets? Thank you!
Alternatives
No response
Additional Content
No response
When i tried to install this controlnet model i get this error in 5.0.0rc2 version what should i do ? Shakker-Labs/FLUX.1-dev-ControlNet-Union-Pro Failed: Unable to determine model type for /workspace/invokeai/models/tmpinstall_y167utiw/FLUX.1-dev-ControlNet-Union-Pro; class FluxControlNetModel is not one of the supported classes [FluxPipeline, StableDiffusionPipeline, StableDiffusionInpaintPipeline, StableDiffusionXLPipeline, StableDiffusionXLImg2ImgPipeline, StableDiffusionXLInpaintPipeline, LatentConsistencyModelPipeline, AutoencoderKL, AutoencoderTiny, ControlNetModel, CLIPVisionModelWithProjection, T2IAdapter, CLIPModel, CLIPTextModel, T5EncoderModel]
Can someone please explain ?
Wait, u guys can even run controlnet with Flux in InvokeAI??? I have 4090 and even with FP8 version of Flux, because invoke will force to load the extra T5 and Clip-I encoder into ur memory, it will occupy the whole 24GB in no time and the generation process will get incrediblely slow already, please see https://github.com/invoke-ai/InvokeAI/issues/6939....
I really dont know how did you manage to run flux in the first place
Well this all wouldnt be an issue if Invoke would know what that my T5 GGUF files were, but instead when i load them or even the safetensors i get "unknown model type"
