ffead92d4e
Revert "Merge pull request #16078 from huchenlei/fix_sd2"
...
This reverts commit 4cc3add77050514ce414
2024-07-06 10:40:48 +03:00
c3ef381cd8
Fix SD2 loading
2024-06-23 11:19:04 -04:00
ebfc9f6d09
Merge branch 'dev' into patch-4
2024-06-08 10:35:09 +03:00
cc9ca67664
Add transformer forward patch
2024-05-17 13:14:26 -04:00
53d67088ee
Patch timestep embedding to create tensor on-device
2024-05-17 12:12:57 -04:00
2a8a60c2c5
Add --precision half cmd option
2024-05-16 19:50:06 -04:00
f0e2098f1a
Add support for `--upcast-sampling` with SD XL
2023-07-18 00:39:50 -04:00
3ba6c3c83c
Fix up string formatting/concatenation to f-strings where feasible
2023-05-09 22:25:39 +03:00
beb7dda5d6
Update sd_hijack_unet.py
2023-03-24 16:25:42 +04:00
91cfa9718c
Update sd_hijack_unet.py
2023-03-21 14:47:43 +04:00
2016733814
Apply hijacks in ddpm_edit for upcast sampling
...
To avoid import errors, ddpm_edit hijacks are done after an instruct pix2pix model is loaded.
2023-02-07 22:53:45 -05:00
ada17dbd7c
Refactor conditional casting, fix upscalers
2023-01-28 04:16:25 -05:00
15e89ef0f6
fix for unet hijack breaking the train tab
2023-01-25 20:11:01 +03:00
84d9ce30cb
Add option for float32 sampling with float16 UNet
...
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().
2023-01-25 01:13:02 -05:00
c0355caefe
Fix various typos
2022-12-14 21:01:32 -05:00
7dbfd8a7d8
do not replace entire unet for the resolution hack
2022-12-10 09:14:45 +03:00
AUTOMATIC1111