KTransformers

Qwen Models

Qwen has both inference tutorials and current SFT candidates. Keep model-family pages precise because Qwen entries vary by method and hardware path.

Inference

ModelEntryMethodStatus
Qwen3-Coder-NextManual SGLang-KT tutorialFP8 or BF16Needs smoke.
Qwen3.5Manual SGLang-KT tutorialBF16 in current tutorial directionNeeds mainline cleanup and smoke.
Qwen3-30B-A3BKT-Kernel README examplesBF16, AMXINT8, LLAMAFILENeeds method-specific smoke.

Fine-Tuning

Use Qwen SFT. Qwen3-235B-A22B and Qwen3.5-397B-A17B examples exist under LLaMA-Factory examples/ktransformers/.

Documentation Source

  • doc/en/kt-kernel/Qwen3-Coder-Next-Tutorial.md
  • doc/en/Qwen3.5.md
  • kt-kernel/README.md
  • LLaMA-Factory examples/ktransformers/train_lora/qwen*.yaml