Local Fine-Tuning
KTransformers fine-tuning extends the same local ownership idea from inference to adaptation. If a workstation can run a MoE model through KTransformers, the project direction is that the same class of machine should be able to train LoRA adapters for it.
System Direction
| Component | Role |
|---|---|
| GPU | Attention, shared paths, residual LoRA capacity, and distributed training control. |
| CPU expert backend | Large MoE expert weights through AMX BF16/INT8/INT4 SFT backends. |
| LLaMA-Factory | User-facing training workflow and dataset/config management. |
| KT integration | use_kt: true, kt_config, and backend-aware expert execution. |
The current public path is LoRA SFT with LLaMA-Factory. This technical page explains the direction; the Fine-Tuning section contains the runnable user docs.