aboutsummaryrefslogtreecommitdiff
path: root/modules/sd_hijack.py
AgeCommit message (Collapse)Author
2022-10-11Add check for psutilbrkirch
2022-10-11Add cross-attention optimization from InvokeAIbrkirch
* Add cross-attention optimization from InvokeAI (~30% speed improvement on MPS) * Add command line option for it * Make it default when CUDA is unavailable
2022-10-11rename hypernetwork dir to hypernetworks to prevent clash with an old ↵AUTOMATIC
filename that people who use zip instead of git clone will have
2022-10-11Merge branch 'master' into hypernetwork-trainingAUTOMATIC
2022-10-11Comma backtrack padding (#2192)hentailord85ez
Comma backtrack padding
2022-10-10allow pascal onwardsC43H66N12O12S2
2022-10-10Add back in output hidden states parameterhentailord85ez
2022-10-10Pad beginning of textual inversion embeddinghentailord85ez
2022-10-10Unlimited Token Workshentailord85ez
Unlimited tokens actually work now. Works with textual inversion too. Replaces the previous not-so-much-working implementation.
2022-10-09Removed unnecessary tmp variableFampai
2022-10-09Updated code for legibilityFampai
2022-10-09Optimized code for Ignoring last CLIP layersFampai
2022-10-08Added ability to ignore last n layers in FrozenCLIPEmbedderFampai
2022-10-08add --force-enable-xformers option and also add messages to console ↵AUTOMATIC
regarding cross attention optimizations
2022-10-08check for ampere without destroying the optimizations. again.C43H66N12O12S2
2022-10-08check for ampereC43H66N12O12S2
2022-10-08why did you do thisAUTOMATIC
2022-10-08restore old opt_split_attention/disable_opt_split_attention logicAUTOMATIC
2022-10-08simplify xfrmers options: --xformers to enable and that's itAUTOMATIC
2022-10-08Merge pull request #1851 from C43H66N12O12S2/flashAUTOMATIC1111
xformers attention
2022-10-08Update sd_hijack.pyC43H66N12O12S2
2022-10-08default to split attention if cuda is available and xformers is notC43H66N12O12S2
2022-10-08fix bug where when using prompt composition, hijack_comments generated ↵MrCheeze
before the final AND will be dropped
2022-10-08fix bugs related to variable prompt lengthsAUTOMATIC
2022-10-08do not let user choose his own prompt token count limitAUTOMATIC
2022-10-08let user choose his own prompt token count limitAUTOMATIC
2022-10-08use new attnblock for xformers pathC43H66N12O12S2
2022-10-08delete broken and unnecessary aliasesC43H66N12O12S2
2022-10-07hypernetwork training mk1AUTOMATIC
2022-10-07make it possible to use hypernetworks without opt split attentionAUTOMATIC
2022-10-07Update sd_hijack.pyC43H66N12O12S2
2022-10-07Update sd_hijack.pyC43H66N12O12S2
2022-10-07Update sd_hijack.pyC43H66N12O12S2
2022-10-07Update sd_hijack.pyC43H66N12O12S2
2022-10-02Merge branch 'master' into stableJairo Correa
2022-10-02fix for incorrect embedding token length calculation (will break seeds that ↵AUTOMATIC
use embeddings, you're welcome!) add option to input initialization text for embeddings
2022-10-02initial support for training textual inversionAUTOMATIC
2022-09-30Merge branch 'master' into fix-vramJairo Correa
2022-09-30add embeddings dirAUTOMATIC
2022-09-29fix for incorrect model weight loading for #814AUTOMATIC
2022-09-29new implementation for attention/emphasisAUTOMATIC
2022-09-29Move silu to sd_hijackJairo Correa
2022-09-27switched the token counter to use hidden buttons instead of api callLiam
2022-09-27added token counter next to txt2img and img2img promptsLiam
2022-09-25potential fix for embeddings no loading on AMD cardsAUTOMATIC
2022-09-25Fix token max lengthguaneec
2022-09-21--opt-split-attention now on by default for torch.cuda, off for others (cpu ↵AUTOMATIC
and MPS; because the option does not work there according to reports)
2022-09-21fix for too large embeddings causing an errorAUTOMATIC
2022-09-20fix a off by one error with embedding at the start of the sentenceAUTOMATIC
2022-09-20add the part that was missing for word textual inversion checksumsAUTOMATIC
generated by cgit v1.2.1 (git 2.18.0) at 2025-05-29 10:36:29 +0000