Commit Graph

1126 Commits

Author SHA1 Message Date
Lincoln Stein
710b908290 Keyboard interrupt retains seed and log information in files produced prior to interrupt. Closes #21 2022-08-23 00:51:38 -04:00
Kevin Gibbons
c80ce4fff5
fix default config to match docs / dream.py 2022-08-22 21:46:22 -07:00
Lincoln Stein
6d1219deec fixed filenames 2022-08-22 23:56:36 -04:00
Lincoln Stein
a20827697c adjusted instructions for the released stable-diffusion-v1 weights 2022-08-22 15:33:27 -04:00
xra
e4eb775b63 added optional parameter to skip subprompt weight normalization
allows more control when fine-tuning
2022-08-23 00:03:32 +09:00
xra
a3632f5b4f improved comments & added warning if value couldn't be parsed correctly 2022-08-22 23:32:01 +09:00
xra
2736d7e15e optional weighting for creative blending of prompts
example: "an apple: a banana:0 a watermelon:0.5"
        the above example turns into 3 sub-prompts:
        "an apple" 1.0 (default if no value)
        "a banana" 0.0
        "a watermelon" 0.5
        The weights are added and normalized
        The resulting image will be: apple 66%, banana 0%, watermelon 33%
2022-08-22 22:59:06 +09:00
Lincoln Stein
aa2729d868 user's prompt is now normalized for reproducibility and written into the destination PNG file as a tEXt metadata chunk named "Dream". You can retrieve the prompt with an image editing program that supports browsing the full metadata, or with the images2prompt.py script located in 'scripts' 2022-08-22 00:12:16 -04:00
Lincoln Stein
5f352aec87 test of normalization of prompt 2022-08-21 22:48:40 -04:00
Lincoln Stein
f5450bad61 k_lms sampling working; half precision working, can override with --full_precision 2022-08-21 20:16:31 -04:00
Lincoln Stein
78aba5b770 preparing for merge into main 2022-08-21 19:57:48 -04:00
Lincoln Stein
bb91ca0462 first attempt to fold k_lms changes proposed by hwharrison and bmaltais 2022-08-21 17:09:00 -04:00
xra
a5fb8469ed use Half precision for reduced memory usage & faster speed
This allows users with 6 & 8gb cards to run 512x512 and for even larger resolutions for bigger GPUs
I compared the output in Beyond Compare and there are minor differences detected at tolerance 3, but side by side the differences are not perceptible.
2022-08-19 17:23:43 +09:00
Lincoln Stein
2b2ebd19e7 fixed a typo that introduced a crash 2022-08-18 13:47:07 -04:00
Lincoln Stein
87fb4186d4 folded in changes from img2img-dev 2022-08-18 12:45:02 -04:00
Lincoln Stein
750408f793 added command-line completion 2022-08-18 12:43:59 -04:00
Lincoln Stein
bf76c4f283 img2img is now working; small refactoring of grid code in simplet2i.py 2022-08-18 10:47:53 -04:00
Lincoln Stein
831bbd7a54 improved error reporting when a missing online dependency can't be downloaded 2022-08-17 18:06:30 -04:00
Lincoln Stein
c477525036 catch and handle malformed user inputs; documentation fixes 2022-08-17 12:35:49 -04:00
Lincoln Stein
a7532b386a simplified instructions to preload Bert and kornia prerequisites; fixed --grid and --batch handling; added timing information after image generation 2022-08-17 12:00:00 -04:00
Lincoln Stein
d6124c44a3 added customized patches and updated the README 2022-08-16 21:34:37 -04:00
Robin Rombach
2ff270f4e0 stable diffusion 2022-08-10 16:30:49 +02:00
rromb
f13bf9bf46 add vqgan loss with codebook statistic eval 2022-02-21 15:06:50 +01:00
Robin Rombach
2b46bcb98c
Update ddpm.py
clean up no.1
2022-01-17 21:24:19 +01:00
ablattmann
171cf29fb5 add configs for training unconditional/class-conditional ldms 2021-12-22 15:57:23 +01:00
ablattmann
e66308c7f2 add code 2021-12-21 03:23:41 +01:00