Hey! Wanted to ask, I have recently trained both a Flux Finetune and LoRA sets following Furkans gui
Hey! Wanted to ask, I have recently trained both a Flux Finetune and LoRA sets following Furkans guides. My data set is about 500 high quality images of interior design featuring different rooms with various materials and furniture. I also have associated natural language captions that were generated with CGPT following instructions on how to describe the images.
The results for the finetune were very lackluster and the LoRA was a bit better, but still not as good as my old SDXL LoRAs in terms of geometric accuracies and textures.
Does anyone have any suggestions on what method I should best pursue? Is Flux the wrong model for this application? I am particularly drawn to its prompt adherence and I additionally intend to use control nets with it.
The results for the finetune were very lackluster and the LoRA was a bit better, but still not as good as my old SDXL LoRAs in terms of geometric accuracies and textures.
Does anyone have any suggestions on what method I should best pursue? Is Flux the wrong model for this application? I am particularly drawn to its prompt adherence and I additionally intend to use control nets with it.
