Skip to content
View original post on X: TinkerOfficial· 30/100AI score30/100

Tinker introduces Dimuon, a LoRA fine-tuning optimizer

AISummary

Tinker has built Dimuon, a new optimizer for LoRA fine-tuning. In its SFT tests, Dimuon often reaches lower loss in fewer steps than AdamW. The company says it is still experimenting with Dimuon and is asking users to try it and share feedback.

Post on XView on X
TinkerVerified on X
@tinkerapi

Dimuon is a new optimizer we built for LoRA fine-tuning. In our SFT tests, it often reaches lower loss in fewer steps than AdamW. We're still experimenting with Dimuon; you can Tinker with it today and share your feedback!

See our optimizers guide at https://tinker-docs.thinkingmachines.ai/tinker/optimizers/

Source: Tinker · x.comPublished

Topics

Deployment & engineeringData & trainingProduct updates