ViperEkura d855c09cf3 fix: use torch.optim.AdamW in ManoAdamW instead of NAdamW ...

- ManoAdamW now uses torch.optim.AdamW(fused=True, betas=(0.9, 0.95)) matching MuonAdamW, eliminating a confounding variable in optimizer comparison experiments
- only NoraNAdamW retains NAdamW, which is correct per the Nora paper design

2026-08-01 09:20:54 +08:00

..

__init__.py

feat: add Mano manifold optimizer (mano_adamw)

2026-08-01 08:51:08 +08:00

composite.py

refactor: extract composite optimizer helpers and unify naming

2026-08-01 08:07:45 +08:00

mano_adamw.py

fix: use torch.optim.AdamW in ManoAdamW instead of NAdamW

2026-08-01 09:20:54 +08:00

muon_adamw.py

refactor: extract composite optimizer helpers and unify naming

2026-08-01 08:07:45 +08:00

nora_nadamw.py

refactor: extract composite optimizer helpers and unify naming

2026-08-01 08:07:45 +08:00