KTransformers

DPO Status

DPO is not part of the current KT fine-tuning support scope yet.

The current fine-tuning section focuses on MoE LoRA SFT through LLaMA-Factory. DPO needs target-environment validation before it can join the current path:

ItemRequired information
Training entryCurrent LLaMA-Factory command; old patching paths are historical references.
KT backendExplicit kt_config backend that maps to current KT SFT code.
ModelExact checkpoint and prepared expert weights if needed.
RuntimeMinimum training validation on the target hardware and software environment.
OutputAdapter files and at least one minimal post-training sanity check.

Until then, older DPO pages remain historical references only.