>>It sure as hell does not have any male assholes in it's training data, so it seems the only way to fix it is training our own models with some gay porn.
>>Using the Google Colab method, it should be an easy task, following youtube tutorials such as :
https://www.youtube.com/watch?v=fyBqkKCIpYU [
Embed]
>>What we need as a starting point is to collect our favorite gay porn pictures as a data set and train the model on those.
>>I think the max number of pics would be around ~80 , more than 100 would prologue the process to more than what the Free tier in google colab allows which is 6 hours so we need to find the balance between the max numbers of pics we can feed it to the time it takes for the process , and make sure it is under 6 hours.
>>Another tip for google colab, since they do random checkups every minute or so , we need to simply do this :
function ConnectButton(){
console.log("Connect pushed");
document.querySelector("#top-toolbar > colab-connect-button").shadowRoot.querySelector("#connect").click()
}
setInterval(ConnectButton,60000);
>right-clicked then inspect the element and then go to console and paste this code
>>This is done in order to bypass the cursed colab minute minder that boots you out in case you're not actively interacting with the tab , so do that .
>Colab direct link :
https://colab.research.google.com/github/TheLastBen/fast-stable-diffusion/blob/main/fast-DreamBooth.ipynb
>>Also do it on the 1.5 base model, using 512x512 pics if you're gonna do it on colab , otherwise it again makes the task much longer which boots you out after 6 hours so remember to keep it under 6 hours.
>>If , once you start the process, you get an approximate estimate that's over 6 hours, cancel it and adjust the number of pictures you feed it, in order to not waste time cose if it goes over the 6 hour mark it will boot you out and you've lost those hours for nothing so best to loose 10 minutes in the beginning to get everything sorted out.
