diff options
| author | Yuren Hao <yurenh2@illinois.edu> | 2026-08-05 04:34:17 -0500 |
|---|---|---|
| committer | Yuren Hao <yurenh2@illinois.edu> | 2026-08-05 04:34:17 -0500 |
| commit | 08f5c8cc3a660da3c2a2ac8af24e3bb221f69a25 (patch) | |
| tree | b3c7f05ab5844934a6255abf220e193b43d2a601 /scripts/bp_transformer.py | |
| parent | 61d55ff1fdf53fd58d8939087ce81f92c7197b7f (diff) | |
优化器审计修订: OLion(sign输出→正交化器精度要求塌到1bit)推翻Muon模拟死刑的前提; 三条未探路线(平衡态正交化/跟踪摊销/局部规则白化); Hyperball=每矩阵一个AGC的最便宜自适应+跨宽深LR迁移; 价目表实验扩臂
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014FAPDWQ49M5Ye3NpTndTpn
Diffstat (limited to 'scripts/bp_transformer.py')
0 files changed, 0 insertions, 0 deletions
