06a3e48237
runpod maybe working ish
2022-11-08 23:00:54 -05:00
60a444ebd1
remove unneeded log
2022-11-07 01:28:02 -05:00
412347b0bf
oopsie bug in logging
2022-11-07 01:26:09 -05:00
ca80fd0126
add option to use .txt file with image filename as caption source
2022-11-07 01:20:43 -05:00
33a79c6dce
prepping txt file reading for captions
2022-11-06 21:28:58 -05:00
e96a2101d2
Merge branch 'multiaspect' into main
2022-11-06 21:26:24 -05:00
853a7500f4
some final cleanup and readme updates
2022-11-06 19:59:37 -05:00
aacbde8bc7
bunch of loader work and optimization stuff for aspect ratio support
2022-11-06 01:25:03 -04:00
b7779591b1
multi aspect ratio support
2022-11-05 11:41:48 -04:00
9ccf8ee212
consistency for code check on ini
2022-11-04 16:17:17 -04:00
99fd2939f3
ignore ini files in training data
2022-11-04 16:14:10 -04:00
88e804b6a7
fix test_dl.py script
2022-11-03 20:30:46 -04:00
89cc295515
multi-aspect image support work
2022-11-03 19:47:54 -04:00
93bbd30caa
cleaning up training process
2022-11-02 22:23:09 -04:00
30f78de976
fix batch size hunting at start of training
2022-10-28 21:37:21 -04:00
cbca4d13ce
fixed shuffle
2022-10-26 02:13:59 -04:00
0ed7b0b007
grr bug in shuffle
2022-10-26 01:58:14 -04:00
f5aba7293d
big readme update, cruft and dead code removed, reverting some stuff back to xavier og
2022-10-25 22:45:47 -04:00
ddb809d2c6
adding everydream training routines
2022-10-22 14:53:01 -04:00
e2462ba2dc
Change dataloader to correctly pair training and regularization images
2022-10-01 13:56:27 +10:00
648b87a200
Modifications to support multiple identifier training
2022-10-01 10:53:24 +10:00
5239367089
Turning off flip.
2022-09-28 16:18:09 -05:00
589f6ea5ed
changed 'demoura' to 'joepenna' for README.md
2022-09-26 18:06:04 -07:00
25e5c2520e
get rid of pycache stuff and make work with windows
2022-09-25 17:09:41 -07:00
ce7d1dffe7
Improvements by David Bielejeski, Niko Pueringer, Maki99
2022-09-24 00:32:50 -07:00
99c2e27053
Tweaks for training people
...
A few tweaks to the repo to help training faces, though I've also used these same settings to train objects and styles.
Also adding a notebook to make it easier to run the code on vast.ai / runpod / other cloud computing platforms.
2022-09-17 16:00:31 -07:00
8f22429d74
initial commit
2022-09-06 00:00:21 -07:00
Victor Hall