We have not added a separate RTX accelerated version for FLUX.1 Krea, but the model is fully compatible with existing FLUX.1 dev codebase. I don't think we made a separate onnx export for it though. Doing 4~8 bit quantized version with SVDQuant would be a nice follow up so that the checkpoint is more friendly for consumer grade hardware.
Comments
We have not added a separate RTX accelerated version for FLUX.1 Krea, but the model is fully compatible with existing FLUX.1 dev codebase. I don't think we made a separate onnx export for it though. Doing 4~8 bit quantized version with SVDQuant would be a nice follow up so that the checkpoint is more friendly for consumer grade hardware.