What is the difference between the 2 ckpt files, one says emaonly?
I see there are 2 ckpt files in the repo https://huggingface.co/runwayml/stable-diffusion-v1-5/tree/main
v1-5-pruned-emaonly.ckptv1-5-pruned.ckpt
Does using v1-5-pruned.ckptgive better results during inferences?
I was hoping to see an answer for this as well :(
v1-5-pruned-emaonly.ckpt : 4.27GB ema-only weight, uses less VRAM - suitable for inference
v1-5-pruned.ckpt : 7.7GB ema+non-ema weights, uses more VRAM - suitable for fine-tuning
The larger model contains more information and more information is always better with visuals, but they made these smaller ones to be for low-end graphics cards/systems that have less then 16gb vmem (give or take)
If the second option has more information then why diffuser's format (spread in folders) weights less then 7.7GB in total?
When training LoRA, it's better to choose v1-5-pruned.ckpt(7.7GB) as the base model? I have a GPU with 16GB VRAM.