- 28 Jan, 2023 1 commit
-
-
brkirch authored
-
- 26 Jan, 2023 3 commits
-
-
brkirch authored
-
AUTOMATIC1111 authored
Fix full previews and--no-half-vae to work correctly with --upcast-sampling
-
brkirch authored
-
- 25 Jan, 2023 20 commits
-
-
AUTOMATIC authored
-
AUTOMATIC authored
add edit_image_conditioning from my earlier edits in case there's an attempt to inegrate pix2pix properly this allows to use pix2pix model in img2img though it won't work well this way
-
AUTOMATIC1111 authored
allow symlinks in the textual inversion embeddings folder
-
AUTOMATIC1111 authored
Update default Mac command line arguments to use --upcast-sampling instead of --no-half
-
brkirch authored
-
AUTOMATIC authored
-
AUTOMATIC authored
-
Alex "mcmonkey" Goodwin authored
-
AUTOMATIC authored
-
AUTOMATIC1111 authored
Add upcast options, full precision sampling from float16 UNet and upcasting attention for inference using SD 2.1 models without --no-half
-
AUTOMATIC1111 authored
Fix extra network thumbs label color
-
AUTOMATIC authored
-
AUTOMATIC1111 authored
Fix Approx NN previews changing first generation result
-
AUTOMATIC1111 authored
Add instruct-pix2pix hijack
-
Kyle authored
Update requirement for transformers to version 4.25.1 to allow instruct-pix2pix demo code to work
-
Kyle authored
Allows loading instruct-pix2pix models via same method as inpainting models in sd_models.py and sd_hijack_ip2p.py Adds ddpm_edit.py necessary for instruct-pix2pix
-
AUTOMATIC1111 authored
Adds X/Y/Z Grid Script
-
AUTOMATIC authored
-
brkirch authored
Adds "Upcast cross attention layer to float32" option in Stable Diffusion settings. This allows for generating images using SD 2.1 models without --no-half or xFormers. In order to make upcasting cross attention layer optimizations possible it is necessary to indent several sections of code in sd_hijack_optimizations.py so that a context manager can be used to disable autocast. Also, even though Stable Diffusion (and Diffusers) only upcast q and k, unfortunately my findings were that most of the cross attention layer optimizations could not function unless v is upcast also.
-
brkirch authored
This also handles type casting so that ROCm and MPS torch devices work correctly without --no-half. One cast is required for deepbooru in deepbooru_model.py, some explicit casting is required for img2img and inpainting. depth_model can't be converted to float16 or it won't work correctly on some systems (it's known to have issues on MPS) so in sd_models.py model.depth_model is removed for model.half().
-
- 24 Jan, 2023 16 commits
-
-
AUTOMATIC authored
-
AUTOMATIC authored
-
AUTOMATIC authored
add PIP_INSTALLER_LOCATION env var to install pip if it's not installed remove accidental call to accelerate when venv is disabled add another env var to skip venv - SKIP_VENV
-
AUTOMATIC authored
-
AUTOMATIC authored
-
AUTOMATIC authored
-
EllangoK authored
-
EllangoK authored
-
EllangoK authored
-
EllangoK authored
this is mostly just so git can detect it properly
-
AUTOMATIC authored
also return the removed field to sdapi/v1/upscalers because someone might have relied on it existing
-
AUTOMATIC authored
-
AUTOMATIC1111 authored
Add exception handling to API image decode
-
AUTOMATIC1111 authored
Make SwinIR upscaler interruptible and skippable
-
AUTOMATIC1111 authored
Set Linux xformers 0.0.16RC425
-
brkirch authored
The loading of the model for approx nn live previews can change the internal state of PyTorch, resulting in a different image. This can be avoided by preloading the approx nn model in advance.
-