Shuai Yang
85e9d294c6
fix a bug for DDIM inversion (DDIMSampler.encode())
...
fix a bug of the incorrect timestep in DDIMSampler.encode() for DDIM inversion
t is incorrectly set as the index of timesteps rather than the timestep in the original code.
2023-03-30 10:09:54 +08:00
Robin Rombach
fa4401ece1
merge unclip into main
2023-03-24 13:02:54 +01:00
Robin Rombach
b4bdae9916
add stable unclip
2023-03-24 11:18:44 +01:00
aalbersk
3349693314
adjust licenses and naming
2023-02-03 09:27:06 +00:00
Anna Alberska
872cc9e873
add info about ninstance and license
2023-01-05 13:16:23 +01:00
aalbersk
7ad54c5ee9
add cpu support & add intel ipex optimizations
2022-12-20 15:18:33 +00:00
Robin Rombach
8bde0cf64f
Merge pull request #89 from Stability-AI/dango.patch.atten_overflow
...
* Force cast to fp32 to avoid atten layer overflow
2022-12-07 14:54:35 +01:00
Dango233
e1797ae248
Add env var for resume previous behavior
2022-12-07 21:38:32 +08:00
Dango233
6e92cda76d
* Force cast to fp32 to avoid atten layer overflow
2022-12-07 19:56:39 +08:00
Ikko Ashimine
d4e8234adf
Fix typo in ddpm.py
...
Paramterization -> Parameterization
2022-11-29 11:52:30 +09:00
Robin Rombach
ca86da3a30
release more models
2022-11-24 01:22:28 +01:00