You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

brkirch e3b53fd295 Add UI setting for upcasting attention to float32 ... 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.		3 years ago
..
api	also return the removed field to sdapi/v1/upscalers because someone might have relied on it existing	3 years ago
codeformer	fix F541 f-string without any placeholders	3 years ago
hypernetworks	enable compact view for train tab	3 years ago
textual_inversion	extra networks UI	3 years ago
call_queue.py	big rework of progressbar/preview system to allow multiple users to prompts at the same time and do not get previews of each other	3 years ago
codeformer_model.py	Set device for facelib/facexlib and gfpgan	3 years ago
deepbooru.py	added an option to filter out deepbooru tags	3 years ago
deepbooru_model.py	Add option for float32 sampling with float16 UNet	3 years ago
devices.py	Add UI setting for upcasting attention to float32	3 years ago
errors.py	better support for xformers flash attention on older versions of torch	3 years ago
esrgan_model.py	Revert "MPS Upscalers Fix"	3 years ago
esrgan_model_arch.py	fix typo in filename for ESRGAN arch	3 years ago
extensions.py	add built-in extension system	3 years ago
extra_networks.py	extra networks UI	3 years ago
extra_networks_hypernet.py	Lora support!	3 years ago
extras.py	third time's the charm	3 years ago
face_restoration.py	codeformer support	3 years ago
generation_parameters_copypaste.py	extra networks UI	3 years ago
gfpgan_model.py	Set device for facelib/facexlib and gfpgan	3 years ago
hashes.py	increase block size	3 years ago
images.py	make it so that PNG images with EXIF do not lose parameters in PNG info tab	3 years ago
img2img.py	bump gradio to 3.16.2	3 years ago
import_hook.py	add a comment for disable xformers hack	3 years ago
interrogate.py	remove fairscale requirement, add fake fairscale to make BLIP not complain about it mk2	3 years ago
localization.py	add ability to create extensions that add localizations	3 years ago
lowvram.py	unload depth model if medvram/lowvram enabled	3 years ago
masking.py	Fix non-square full resolution inpainting.	3 years ago
memmon.py	add additional memory states	3 years ago
modelloader.py	add more stuff to ignore when creating model from config	3 years ago
ngrok.py	Correct singleton comparisons	3 years ago
paths.py	fix BLIP failing to import depending on configuration	3 years ago
postprocessing.py	a possible fix for broken image upscaling	3 years ago
processing.py	Add UI setting for upcasting attention to float32	3 years ago
progress.py	add progress bar to modelmerger	3 years ago
prompt_parser.py	add BREAK keyword to end current text chunk and start the next	3 years ago
realesrgan_model.py	Fix unable to find Real-ESRGAN model info error (AttributeError: 'NoneType' object has no attribute 'data_path') #6841 #5170	3 years ago
safe.py	Merge pull request #5992 from yuvalabou/F541	3 years ago
script_callbacks.py	Lora support!	3 years ago
script_loading.py	make existing script loading and new preload code use same code for loading modules	3 years ago
scripts.py	rework extras tab to use script system	3 years ago
scripts_postprocessing.py	rework extras tab to use script system	3 years ago
sd_disable_initialization.py	fix missing field for aesthetic embedding extension	3 years ago
sd_hijack.py	write a comment for fix_checkpoint function	3 years ago
sd_hijack_checkpoint.py	add option to show/hide warnings	3 years ago
sd_hijack_clip.py	add BREAK keyword to end current text chunk and start the next	3 years ago
sd_hijack_clip_old.py	CLIP hijack rework	3 years ago
sd_hijack_inpainting.py	find configs for models at runtime rather than when starting	3 years ago
sd_hijack_open_clip.py	Add support Stable Diffusion 2.0	3 years ago
sd_hijack_optimizations.py	Add UI setting for upcasting attention to float32	3 years ago
sd_hijack_unet.py	Add option for float32 sampling with float16 UNet	3 years ago
sd_hijack_utils.py	Add option for float32 sampling with float16 UNet	3 years ago
sd_hijack_xlmr.py	alt-diffusion integration	3 years ago
sd_models.py	Add option for float32 sampling with float16 UNet	3 years ago
sd_samplers.py	Add a check and explanation for tensor with all NaNs.	3 years ago
sd_vae.py	allow baking in VAE in checkpoint merger tab	3 years ago
sd_vae_approx.py	Fix Approx NN on devices other than CUDA	3 years ago
shared.py	Add UI setting for upcasting attention to float32	3 years ago
styles.py	change style dropdowns to multiselect	3 years ago
sub_quadratic_attention.py	Add UI setting for upcasting attention to float32	3 years ago
txt2img.py	bump gradio to 3.16.2	3 years ago
ui.py	fix open directory button failing	3 years ago
ui_common.py	fix open directory button failing	3 years ago
ui_components.py	rework extras tab to use script system	3 years ago
ui_extensions.py	sort extensions by date and add an option to sort by other columns	3 years ago
ui_extra_networks.py	add support for apostrophe in extra network names	3 years ago
ui_extra_networks_hypernets.py	add a slider for default value of added extra networks	3 years ago
ui_extra_networks_textual_inversion.py	fix broken textual inversion extras tab	3 years ago
ui_postprocessing.py	rework extras tab to use script system	3 years ago
ui_tempdir.py	add support for running with gradio 3.9 installed	3 years ago
upscaler.py	Fix unable to find Real-ESRGAN model info error (AttributeError: 'NoneType' object has no attribute 'data_path') #6841 #5170	3 years ago
xlmr.py	fix bugs	3 years ago