From f1aa1d67118f149b081411d74ebfe7b311cd169f Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Thu, 1 Sep 2022 18:19:39 +0300 Subject: [PATCH] clarifications for --lowvram and --medvram in readme. --- README.md | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) diff --git a/README.md b/README.md index f2635d0a..6e1c7656 100644 --- a/README.md +++ b/README.md @@ -294,11 +294,21 @@ Model is separated into modules, and only one module is kept in GPU memory; when is removed from GPU memory. The nature of this optimization makes the processing run slower -- about 10 times slower compared to normal operation on my RTX 3090. -`--medvram` is another optimization that should reduce VRAM usage significantly by not peocessing conditional and +`--medvram` is another optimization that should reduce VRAM usage significantly by not processing conditional and unconditional denoising in a same batch. This implementation of optimization does not require any modification to original Stable Diffusion code. +#### What option to use? +If you have 4GB VRAM and want to make 512x512 (or maybe up to 640x640) images, use `--medvram`. + +If you have 4GB VRAM and want to make images larger than that, use `--lowvram`. + +If you have more VRAM and want to make larger images than you can usually make, use `--medvram`. You can use `--lowvram` +also but the effect will likely be barely noticeable. + +Otherwise, do not use any. + ### Inpainting In img2img tab, draw a mask over a part of image, and that part will be in-painted.