All of my own trained loras just output black

I am using https://github.com/bmaltais/kohya_ss to train my loras with these settings https://www.mediafire.com/file/p7hc4q8n7wuu31z/CharacterLoraSettings.json/file

I am using https://github.com/vladmandic/automatic as my stable diffusion because https://github.com/AUTOMATIC1111/stable-diffusion-webui kept giving me error issues and a reddit thread said to switch to this fork and it would fix my error issues and it did. Any other lora works just fine just not the ones that i train. My training images are all 512 x 512 like i have in the settings. I have used both stable-diffusion-v1-5 and stable-diffusion-2-1-base as my basses with the same outcome. I am running it on a new 3060 12gb so i know that it is not a gpu related issue. The sample images that it gives me are fine just not the output in stable diffusion they are all black. Any help would be greatly appreciated

Start by reproducing the black-output result with the linked CharacterLoraSettings.json and the reported Stable Diffusion base models. Compare the training sample images with inference output and document whether the cause is in training or in the external inference setup; done means identifying a reproducible cause or a clear missing configuration detail.

Written by the indexing model from the issue text.