- Bugs / crashes → bug report.
- ROCm / gfx compatibility → ROCm compatibility (include GPU, gfx arch, ROCm, torch, llama.cpp/vLLM commit).
- Score / benchmark questions → benchmark report.
- Security → see SECURITY.md (private reporting only).
- Usage questions / discussion → GitHub Discussions (if enabled) or open a generic issue.
Before filing, run hunyuan-ocr doctor and paste its output — it captures your
environment without loading the model.
This is a community project; responses are best-effort and not on an SLA.