Commit graph

48 commits

Author SHA1 Message Date
rromb
6663c9bb71 add alternative guidance strategy 2022-06-13 23:35:05 +02:00
Patrick Esser
6f604acf0a never run out of data 2022-06-13 10:59:50 +00:00
Patrick Esser
046410167f solve interpretability 2022-06-13 10:56:17 +00:00
Patrick Esser
3c77204c0a experimental support for guided upscale sampling 2022-06-13 10:44:10 +00:00
rromb
255d479088 add logging for upscaler 2022-06-13 10:43:41 +02:00
Patrick Esser
c89452ef9a support lr creation in laion 2022-06-12 23:26:31 +00:00
rromb
c5a39aff8a add upscaling 2022-06-13 00:39:48 +02:00
Patrick Esser
250f89c1a3 initial code to try and resume larger models from smaller models 2022-06-11 18:35:49 -04:00
rromb
5e3f9cec53 fix it 2022-06-10 11:44:55 +02:00
rromb
cbbef13b7d fix it 2022-06-10 11:43:49 +02:00
rromb
18a9202196 f16 with size filtering 2022-06-10 11:08:42 +02:00
rromb
b6cf25386a sampling updates 2022-06-10 10:54:14 +02:00
rromb
8d8369a3d5 Merge branch 'main' of github.com:pesser/stable-diffusion into main 2022-06-09 10:57:07 +02:00
rromb
a133096ddf coco 2022-06-09 10:56:41 +02:00
rromb
9cda7feafa eval stuff 2022-06-09 10:56:34 +02:00
root
5d40de00fb Merge remote-tracking branch 'github/main' into main 2022-06-08 21:04:40 +00:00
root
58f2ba8a44 512 config 2022-06-08 21:00:53 +00:00
rromb
a0f390afb6 dynamic thresholding (maybe?) 2022-06-05 19:22:07 +02:00
rromb
64cc718ead more specific dummy data 2022-06-05 19:21:22 +02:00
Patrick Esser
7d432123d5 filter examples without req keys, make collation more robust 2022-06-01 22:04:45 +00:00
Patrick Esser
cb0414d89b goooo 2022-06-01 09:14:09 +00:00
root
3769989f20 Merge remote-tracking branch 'github/main' into main 2022-06-01 08:51:55 +00:00
Patrick Esser
2cbbb3148a optional use of ema 2022-06-01 08:46:11 +00:00
rromb
fff19bf82e handle listconfig 2022-06-01 09:52:17 +02:00
rromb
b3a604d870 add cfg to log_images 2022-06-01 09:36:48 +02:00
rromb
abd9fb7af1 handle it 2022-05-31 22:54:59 +02:00
rromb
fdc93bd76d fix import 2022-05-31 15:56:29 +02:00
rromb
de7a003484 one more handler 2022-05-31 15:21:55 +02:00
rromb
1eb0b88a8d add alstroemeria data 2022-05-31 15:20:27 +02:00
rromb
07ebb8fb2a disabled_train is set later 2022-05-31 14:18:01 +02:00
rromb
17f4ca8110 fix clip encoder code 2022-05-31 12:36:26 +02:00
rromb
e637491838 clip encoder 2022-05-31 12:28:00 +02:00
rromb
3c53c6c15b add frozen t5 encoder from huggingface 2022-05-31 11:42:53 +02:00
Patrick Esser
ac1ca73d69 simplify data loader 2022-05-30 20:35:27 +00:00
pesser
74797d4099 warn and continue 2022-05-30 14:09:27 +00:00
pesser
f8c559e862 ignore 2022-05-30 14:04:35 +00:00
pesser
3e2e9d8e9e debug with dummy data 2022-05-30 13:09:42 +00:00
rromb
9a419a1b14 first (in)stable steps 2022-05-27 11:46:04 +02:00
pesser
f7a6152022 laion webdataset example code 2022-05-26 21:51:29 +00:00
Robin Rombach
4c8bff393a
Merge pull request #51 from crowsonkb/main
Add PLMS sampling and do one 2x size batch per sampling step
2022-04-15 17:35:51 +02:00
rromb
71ebe40791 do not fix num_heads to one in legacy mode 2022-04-15 17:24:11 +02:00
Katherine Crowson
66df437e52 Combine the two classifier-free guidance model outputs into a single batch 2022-04-05 11:35:05 -07:00
Katherine Crowson
f0c4e092c1 Add PLMS sampling 2022-04-04 14:45:35 -07:00
Robin Rombach
32c525011b
unconditional guidance with general s!=1.0
Example: "negative weights"
2022-04-04 22:31:35 +02:00
rromb
a80ed7987e add new models 2022-04-04 16:17:48 +02:00
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