torch.manual_seed() already sets a CPU generator, so there is no reason to create a CPU generator manually. torch.randn_like also needs a MPS fix for k-diffusion, but a torch hijack with randn_like already exists so it can also be used for that.

fix an error that happens when you type into prompt while switching model, put queue stuff into separate file

if image on disk was deleted between being generated and request being completed, do use temporary dir to store it for the browser

resolve [name] after resolving [filewords] in training

Add CORS-allow policy launch argument using regex

Move progress info to beginning of site title

Fix support for devices other than CUDA in DeepBooru

Fix api ignoring sampler_name settings

CI tests with github-actions and some improvements to testing

fix null negative_prompt on get requests

Forcing HTTPS instead of HTTP for ngrok

rework #5012 to also work for pictures dragged into the prompt and also add Clip skip + ENSD to parameters

Support NAI style exif in PNG Info for Send... buttons

Fix API img2img not accepting bare base64 strings

Supporting `*.safetensors` format.

Remove cmd args requirement for deepdanbooru

Added a flag to run as root if needed

fix the model name error of Real-ESRGAN in the opts default value

add model_name pattern for saving

no-half support for SD 2.0

Fixed issue with selected_gallery_index()

#3449 - VRAM leak when switching to/from inpainting checkpoint

Fix setting device for GFPGAN and CodeFormer

Fixes for PyTorch 1.12.1 when using MPS

torch.cuda.empty_cache() defaults to cuda:0 device unless explicitly …

Add DPM++ SDE sampler