Jona Reynaert
|
d0e40db648
|
fix: adapted checkpoint eval to handle decentralized models
|
2026-05-09 15:54:35 +02:00 |
|
Jona Reynaert
|
2594c53d49
|
Merge branch 'dev' into feat/message_passing
|
2026-05-09 15:02:25 +02:00 |
|
Robin Meersman
|
a0225eb48c
|
test(obs processor): general setup for tests
|
2026-05-09 13:31:19 +02:00 |
|
Robin Meersman
|
6fc4f752a8
|
fix(my fuckery): removed shit that did not need to be in the repo :)
|
2026-05-08 23:51:03 +02:00 |
|
Robin Meersman
|
c6490eb672
|
Merge branch 'feat/message_passing' of github.com:SELab-3-2026/SEL3-2026-Groep-4 into feat/message_passing
|
2026-05-08 23:49:45 +02:00 |
|
Robin Meersman
|
17666ed5ba
|
feat(obs_processor): rewrote split method, todo: testing + code cleanup
|
2026-05-08 23:49:36 +02:00 |
|
Robin Meersman
|
34f0b1b80e
|
feat(obs_processor): rewrote split method, todo: testing + code cleanup
|
2026-05-08 23:46:45 +02:00 |
|
|
|
fddf46c799
|
Merge pull request #53 from SELab-3-2026/feat/training-evaluation
|
2026-05-07 23:13:34 +02:00 |
|
|
|
cf6751ad15
|
Merge branch 'dev' into feat/training-evaluation
|
2026-05-07 23:11:25 +02:00 |
|
|
|
1f14ccfd4a
|
refactor: use mjx eval in PPOTrainer
|
2026-05-07 23:01:56 +02:00 |
|
|
|
a7db4c77bb
|
refactor(log): extract w&b file sync
|
2026-05-07 22:50:38 +02:00 |
|
|
|
14d719fe74
|
refactor: extract mjx evaluation
|
2026-05-07 22:48:23 +02:00 |
|
Robin Meersman
|
07ee32fd3b
|
fix(simulate): actor action dimension updated to new way of calculating (as in training)
|
2026-05-07 19:53:49 +02:00 |
|
|
|
6956c5e853
|
refactor: cleanups while reviewing
|
2026-05-07 19:08:18 +02:00 |
|
Robin Meersman
|
4e3cd0c673
|
fix(ppo): identity function now takes 2 params, like message_passer
|
2026-05-07 14:07:19 +02:00 |
|
Robin Meersman
|
517c3134ca
|
fix: rewrote info logger calls to debug calls for debug related logs
|
2026-05-07 14:00:39 +02:00 |
|
RobinMeersman
|
4c5122e779
|
Update tests/test_adjacency.py
Co-authored-by: Tibo De Peuter <tibo.depeuter@ugent.be>
|
2026-05-07 13:57:19 +02:00 |
|
RobinMeersman
|
7026125bba
|
Update scripts/hpc/train.pbs
Co-authored-by: Tibo De Peuter <tibo.depeuter@ugent.be>
|
2026-05-07 13:56:28 +02:00 |
|
Robin Meersman
|
7f5b730cdf
|
fix(ppo trainer): added check if message passing is None
|
2026-05-07 13:52:57 +02:00 |
|
Jona Reynaert
|
b91908bf7a
|
feat: abstracted checkpoint eval configs
|
2026-05-07 13:45:03 +02:00 |
|
Jona Reynaert
|
67c459ca94
|
fix: changed checkpoint evaluation csv fields
|
2026-05-07 12:03:53 +02:00 |
|
Robin Meersman
|
23fcee2110
|
fix(ppo training): forgot extra change from dev
|
2026-05-07 08:42:35 +02:00 |
|
RobinMeersman
|
090e774ad7
|
Merge branch 'dev' into feat/message_passing
|
2026-05-07 08:37:30 +02:00 |
|
Cedric
|
4068ccef5d
|
fix: split pad dada
|
2026-05-06 21:51:39 +00:00 |
|
Cedric
|
9c20c28b2a
|
feat: split n flat n pad decentered
|
2026-05-06 18:52:19 +00:00 |
|
Robin Meersman
|
5e7fd27b61
|
backup
|
2026-05-06 19:26:32 +02:00 |
|
Jona Reynaert
|
51d232b7f6
|
Merge pull request #52 from SELab-3-2026/feat/reset-training-envs
feat: reset training envs when goal is reached
|
2026-05-06 17:30:41 +02:00 |
|
Robin Meersman
|
af68a0064c
|
fix(ppo): added message_passer to ppo
|
2026-05-05 18:22:52 +02:00 |
|
Robin Meersman
|
079480324d
|
fix(testing): forward pass test updated to new abstraction
|
2026-05-05 12:11:37 +02:00 |
|
Robin Meersman
|
3ad1ad8e39
|
fix(testing): target direction updated to work with dim >= 2
|
2026-05-05 11:49:32 +02:00 |
|
Robin Meersman
|
b244761db1
|
cleanup(MLPs): moved build adjacency to own file, made MLPs a full python package
|
2026-05-04 22:39:33 +02:00 |
|
Robin Meersman
|
2ac21e2014
|
fix: linting + format
|
2026-05-04 22:38:45 +02:00 |
|
Robin Meersman
|
43de6a9670
|
merge: merged dev into branch
|
2026-05-04 22:23:58 +02:00 |
|
RobinMeersman
|
61064ca70e
|
Merge branch 'dev' into feat/message_passing
|
2026-05-04 21:44:06 +02:00 |
|
Robin Meersman
|
791dcf231c
|
other(ppo training loop): code cleanup + enabled debug logging of model parameters
|
2026-05-04 21:29:27 +02:00 |
|
Robin Meersman
|
dd2612e2ec
|
fix(ppo training): more shape-related bug fixes + moved computation of #arms and segments outside of jit compiled function
|
2026-05-04 21:26:34 +02:00 |
|
Robin Meersman
|
a43f3c0172
|
fix(ppo training): fixed shape error while clipping action
|
2026-05-04 19:36:59 +02:00 |
|
Jona Reynaert
|
0a00a978e8
|
fix: formatted file
|
2026-05-04 00:00:05 +02:00 |
|
Jona Reynaert
|
f66e073f18
|
feat: reset training envs when goal is reached
|
2026-05-03 23:56:27 +02:00 |
|
Jona Reynaert
|
0b845dd7a9
|
fix: improved speed of checkpoint eval
|
2026-05-03 18:26:24 +02:00 |
|
Jona Reynaert
|
a3d6b32eaa
|
feat: implemented checkpoint evaluation
|
2026-05-03 15:53:47 +02:00 |
|
Jona Reynaert
|
90c2c3d1f0
|
feat: added flags for checkpoint evaluation
|
2026-05-03 15:53:26 +02:00 |
|
Robin Meersman
|
bb14507fc7
|
fix(message passing): forgot vmapping over params :thumbs_up:
|
2026-05-02 23:34:52 +02:00 |
|
Robin Meersman
|
3cb3c765cc
|
feat(message passing): fixed tracing bug, todo: dense layer input dimension incorrect
|
2026-05-02 23:27:31 +02:00 |
|
Robin Meersman
|
027b13b30f
|
feat(message passing): used config value for message passing repetitions
|
2026-05-02 23:15:55 +02:00 |
|
Robin Meersman
|
6114e36e27
|
feat(PPOTrainer): decentralized message passing babyyyy
|
2026-05-02 21:50:42 +02:00 |
|
|
|
70ef9d4058
|
Merge pull request #45 from SELab-3-2026/fix/input-space
|
2026-05-02 20:30:59 +02:00 |
|
JibrilExe
|
9398592e34
|
merged
|
2026-05-02 11:12:01 +02:00 |
|
JibrilExe
|
2b46b3f5d2
|
fix: only flatten actions for environ
|
2026-05-02 11:08:48 +02:00 |
|
JibrilExe
|
f77766ba08
|
fix:logging shapes
|
2026-05-02 09:38:22 +02:00 |
|
