Just did a thousand images (different seeds for the same prompt) with the base (not fork) model at 50 iterations in a couple of hours (RTX 3080). Gpu vram usage

Just did a thousand images (different seeds for the same prompt) with the base (not fork) model at 50 iterations in a couple of hours (RTX 3080). Gpu vram usage seemed to hover around 9 gigs. This model is pretty damn fast for what it does. I quickly threw together a folder structure where I have a md5'd prompt as a folder name, into that goes _promp.txt with the actual text of the prompt and the images i generate in a loop with the seed used and iteration number in the image's file name. That way I can generate like 20 seed-based images for a prompt and if the model bites I let it run with a much higher number of seed-based images. When you have a 1000 to pick from, some of the results are freaking amazing.