diff options
| author | YurenHao0426 <Blackhao0426@gmail.com> | 2026-03-23 19:46:56 -0500 |
|---|---|---|
| committer | YurenHao0426 <Blackhao0426@gmail.com> | 2026-03-23 19:46:56 -0500 |
| commit | 3012cba6032ee04cc0b82c178fbf8df8e47c7d2f (patch) | |
| tree | f78b2cfb1202a7dd08f9ddab08cb122e5264099d /experiments/cifar_resmlp.py | |
| parent | 32123cb36ae9521f60c9b6f67458b931b6540ef2 (diff) | |
Add sweep results confirming terminal gradient matching is essential
12-config sweep: no hyperparameter combination recovers useful credit
gradients without terminal gradient matching (best cos ~0.3 early, decays to ~0).
Diffstat (limited to 'experiments/cifar_resmlp.py')
0 files changed, 0 insertions, 0 deletions