aboutsummaryrefslogtreecommitdiff
path: root/modules/processing.py
AgeCommit message (Collapse)Author
2023-01-29Extra network in hr abomination fixinvincibledude
2023-01-29Extra networks loading fixinvincibledude
2023-01-29Extra networks loading fixinvincibledude
2023-01-29Extra networks loading fixinvincibledude
2023-01-29Merge branch 'master' into masterInvincibleDude
2023-01-29remove Batch size and Batch pos from textinfo (goodbye)AUTOMATIC
2023-01-28Merge pull request #7309 from brkirch/fix-embeddingsAUTOMATIC1111
Fix embeddings, upscalers, and refactor `--upcast-sampling`
2023-01-28Refactor conditional casting, fix upscalersbrkirch
2023-01-27add data-dir flag and set all user data directories based on itMax Audron
2023-01-26add an option to enable sections from extras tab in txt2img/img2imgAUTOMATIC
fix some style inconsistenices
2023-01-26Fix full previews, --no-half-vaebrkirch
2023-01-25add edit_image_conditioning from my earlier edits in case there's an attempt ↵AUTOMATIC
to inegrate pix2pix properly this allows to use pix2pix model in img2img though it won't work well this way
2023-01-25Merge pull request #6510 from brkirch/unet16-upcast-precisionAUTOMATIC1111
Add upcast options, full precision sampling from float16 UNet and upcasting attention for inference using SD 2.1 models without --no-half
2023-01-25change to code for live preview fix on OSX to be bit more obviousAUTOMATIC
2023-01-25Add UI setting for upcasting attention to float32brkirch
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.
2023-01-25Add option for float32 sampling with float16 UNetbrkirch
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().
2023-01-24Merge branch 'AUTOMATIC1111:master' into masterInvincibleDude
2023-01-23Fix different first gen with Approx NN previewsbrkirch
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.
2023-01-22Gen params paste improvementinvincibledude
2023-01-22Gen params paste improvementinvincibledude
2023-01-22UI and PNG info improvementsinvincibledude
2023-01-22UI and PNG info improvementsinvincibledude
2023-01-22hr conditioninginvincibledude
2023-01-22hr conditioninginvincibledude
2023-01-22hr conditioninginvincibledude
2023-01-22hr conditioninginvincibledude
2023-01-22hr conditioninginvincibledude
2023-01-22hr conditioninginvincibledude
2023-01-22Hr-fix separate prompt experimentationinvincibledude
2023-01-22Logging for debugginginvincibledude
2023-01-22Fixinvincibledude
2023-01-22Hr separate prompt testinvincibledude
2023-01-22PLMS edge-case handling fix 5invincibledude
2023-01-22PLMS edge-case handling fix 3invincibledude
2023-01-22PLMS edge-case handling fix 2invincibledude
2023-01-22PLMS edge-case handling fixinvincibledude
2023-01-22enable compact view for train tabAUTOMATIC
prevent previews from ruining hypernetwork training
2023-01-21Type mismatch fixinvincibledude
2023-01-21First test of different sampler for hi-res fixinvincibledude
2023-01-21extract extra network data from prompt earlierAUTOMATIC
2023-01-21make it so that extra networks are not removed from infotextAUTOMATIC
2023-01-21extra networks UIAUTOMATIC
rework of hypernets: rather than via settings, hypernets are added directly to prompt as <hypernet:name:weight>
2023-01-18Merge pull request #6854 from EllangoK/masterAUTOMATIC1111
Saves Extra Generation Parameters to params.txt
2023-01-18use DDIM in hires fix is the sampler is PLMSAUTOMATIC
2023-01-17Changed params.txt save to after manual init callEllangoK
2023-01-16make StableDiffusionProcessing class not hold a reference to shared.sd_model ↵AUTOMATIC
object
2023-01-16Add a check and explanation for tensor with all NaNs.AUTOMATIC
2023-01-14change hypernets to use sha256 hashesAUTOMATIC
2023-01-12Fix extension parameters not being saved to last used parametersspace-nuko
2023-01-09add an option to use old hiresfix width/height behaviorAUTOMATIC
add a visual effect to inactive hires fix elements