Commit Graph

33 Commits

Author SHA1 Message Date
Michael Pilosov, PhD
120175b2d9 dont need this many animations 2024-02-23 16:45:32 +00:00
Michael Pilosov, PhD
d72608756b really high res images... will be slow. 2024-02-22 08:28:40 +00:00
Michael Pilosov, PhD
55f0f60bb8 prepare animations (supervised and not) 2024-02-22 07:49:54 +00:00
Michael Pilosov, PhD
d330d9b4ca supervised experiment 2024-02-22 07:44:52 +00:00
Michael Pilosov, PhD
0dc1927d2e linting 2024-02-22 07:37:33 +00:00
Michael Pilosov, PhD
5def982f12 recreate iris looking result. weird though. 2024-01-28 08:09:13 +00:00
Michael Pilosov, PhD
a9e772f34e restrict experiment, add loop option + dropout 2024-01-28 07:31:49 +00:00
Michael Pilosov
697030b7df partial unsupervision 2024-01-27 23:34:25 -07:00
Michael Pilosov, PhD
0c4990e98c remove depth = 8 and switch to relu 2024-01-28 05:42:04 +00:00
Michael Pilosov, PhD
1c116f3f12 benchmark supervised again 2024-01-28 02:54:30 +00:00
Michael Pilosov, PhD
c7ffd09fb4 lr scheduler 2024-01-28 02:44:36 +00:00
Michael Pilosov, PhD
4342a54cc8 anchors and more train data, exp w batch 2024-01-28 02:43:35 +00:00
Michael Pilosov, PhD
9e4861a272 try unsupervised again, but with 10k random samples 2024-01-28 01:33:10 +00:00
Michael Pilosov, PhD
e5b6f287a3 xkcd colors may be too few to learn from. need 10x 2024-01-28 01:32:56 +00:00
Michael Pilosov, PhD
865e7f5104 supervised questionable 2024-01-28 01:25:10 +00:00
Michael Pilosov, PhD
953488be4c remove modulo from 3-space 2024-01-27 23:52:04 +00:00
Michael Pilosov, PhD
0e561aae4c big batch 2024-01-27 22:38:01 +00:00
Michael Pilosov, PhD
30470f13bc unsupervised results looking stunning 2024-01-27 22:26:41 +00:00
Michael Pilosov, PhD
a49f166252 seems to learn (supervised) consistently now 2024-01-27 21:55:45 +00:00
Michael Pilosov, PhD
a44580a15b depth of 1 led to consistently bad learning 2024-01-27 21:50:33 +00:00
Michael Pilosov, PhD
6260e7fdd3 restrict optimizers 2024-01-27 21:08:44 +00:00
Michael Pilosov, PhD
467b3f7e57 entirely supervised 2024-01-27 21:03:32 +00:00
Michael Pilosov, PhD
b5d9e725b3 allow for mix of supervised and not with alpha 2024-01-27 19:48:04 +00:00
Michael Pilosov, PhD
b9d334e49a weird setup, but sometimes possible to learn labels 2024-01-27 19:38:40 +00:00
Michael Pilosov, PhD
07b4e548e2 lets make sure this can learn via supervision first 2024-01-27 18:58:06 +00:00
Michael Pilosov, PhD
948bc31861 cull experimental range 2024-01-27 17:58:36 +00:00
Michael Pilosov, PhD
fff2b88fa1 job updates 2024-01-27 10:14:52 +00:00
Michael Pilosov, PhD
49e6260346 experiment loggin setup 2024-01-27 09:50:20 +00:00
Michael Pilosov, PhD
6d40d39097 save config to out 2024-01-27 09:39:25 +00:00
Michael Pilosov, PhD
3b700aee70 optimizer search 2024-01-27 09:34:10 +00:00
Michael Pilosov, PhD
1e818aa977 remove s loss entirely 2024-01-27 09:29:22 +00:00
Michael Pilosov, PhD
1ea29ba11e some missing params 2024-01-27 09:26:19 +00:00
Michael Pilosov, PhD
e1ac3211b9 use lightning CLI everywhere 2024-01-27 09:14:14 +00:00