KTransformers

Troubleshooting

Start with:

kt version
kt doctor

Install Issues

SymptomCheck
kt command not foundConfirm the environment where kt-kernel was installed is active.
Import or wheel errorConfirm Python version and Linux x86-64 environment.
CUDA-related failureConfirm driver, PyTorch CUDA variant, and GPU architecture.
AMX kernel not selectedCheck lscpu flags and BIOS/kernel AMX support.

Serving Issues

SymptomCheck
Server starts but model loading failsVerify --model-path, --kt-weight-path, and --kt-method match.
Output format is unexpectedCheck chat template, parser options, and served model name.
OOM during startupLower GPU expert count or token limits; confirm model-specific memory assumptions.
Slow prefillCheck method, CPU backend, NUMA settings, and layerwise prefill threshold.

Fine-Tuning Issues

SymptomCheck
KT backend not enabledConfirm use_kt: true in training YAML and kt_config.enabled: true in Accelerate config.
Backend mismatchMatch kt_backend with BF16 or converted INT8/INT4 expert weights.
LLaMA-Factory cannot import KT packagesConfirm pip install -r requirements/ktransformers.txt was run in the LLaMA-Factory environment.

Escalation Data

When filing an issue, include the runtime tuple, full launch command, package versions, hardware summary, and the first blocking log lines.