How can I fine tune Realistic_Vision_V6.0_B1_noVAE model and generate my own images? Here is the huggingface link.
I have used autotrain on stable diffusion base xl model. But using it on this Realistic Vision model seems not quite right. Model generates bad images.
autotrain dreambooth --model SG161222/Realistic_Vision_V6.0_B1_noVAE --image-path input_images/ --prompt "photo of maha person" --resolution 1024 --batch-size 1 --num-steps 500 --mixed-precision fp16 --gradient-accumulation 4 --lr 1e-4 --project-name realistic-vision
this gives me pytorch_lora_weights.safetensors file and I use this to generate my own images. pipe.load_lora_weights("model/", weight_name="pytorch_lora_weights.safetensors")
Is this correct technique of fine-tuning Realistic_Vision_V6.0_B1_noVAE model? Or any model like stable diffusion?