 AUTOMATIC1111
|
7e5cdaab4b
SD3 lora support
|
il y a 1 an |
|
AUTOMATIC1111
|
9e404c3154
fix --medvram
|
il y a 1 an |
|
AUTOMATIC1111
|
7e4b06fcd0
support loading clip/t5 from the main model checkpoint
|
il y a 1 an |
|
AUTOMATIC1111
|
d67348a0a5
allow generation to be started with any dimensions specified
|
il y a 1 an |
|
AUTOMATIC1111
|
179ae47d64
fix the problem with infinite prompts where empty cond would be calculated incorrectly
|
il y a 1 an |
|
AUTOMATIC1111
|
0b64633584
fix img2img
|
il y a 1 an |
|
AUTOMATIC1111
|
0c7bdcc1b1
add the missing get_first_stage_encoding function
|
il y a 1 an |
|
AUTOMATIC1111
|
d686e73daa
support for SD3: infinite prompt length, token counting
|
il y a 1 an |
|
AUTOMATIC1111
|
a8fba9af35
medvram support for SD3
|
il y a 1 an |
|
AUTOMATIC1111
|
34b4443cc3
add an option (on by default) to disable T5
|
il y a 1 an |
|
AUTOMATIC1111
|
d4b814aed6
change t5xxl checkpoint to fp8
|
il y a 1 an |
|
AUTOMATIC1111
|
58dc35a64a
change CLIP links to allow anonymous downloading
|
il y a 1 an |
|
AUTOMATIC1111
|
06d0a5ab4d
fix NaN issue when running without --precision half
|
il y a 1 an |
|
AUTOMATIC1111
|
b443fdcf76
prevent accidental creation of CLIP models in float32 type when user wants float16
|
il y a 1 an |
|
AUTOMATIC1111
|
5b2a60b8e2
initial SD3 support
|
il y a 1 an |
