vai-quantization-guide
Master orchestrator for mixed-precision ONNX model quantization targeting AMD NPU (Strix/Telluride). Guides users through quantization strategy selection, accuracy validation, compiler flag configuration, and board execution. Supports full bf16, full vint8, hybrid BF16/VINT8 configurations, and edge quantization in runtime. Specialized quantization methods are provided as optional plugin skills discovered at runtime.
- Version
- 6.3.0
- License
- MIT
- Compatibility
- Requires VitisAI 6.3+
Pinned to revision 2cdc9eef1b6b, so it is the text this page describes rather than whatever the author pushed since.
Files
- skills/vai-quantization-guide/SKILL.md
- skills/vai-quantization-guide/README.md
- skills/vai-quantization-guide/scripts/compare_accuracy.py
- skills/vai-quantization-guide/scripts/dequantize_model.py
- skills/vai-quantization-guide/scripts/extract_ffns.py
- skills/vai-quantization-guide/scripts/ffn_quantize.py
- skills/vai-quantization-guide/scripts/mixed_precision_quantize.py
- skills/vai-quantization-guide/scripts/patch_model_for_vaip.py
- skills/vai-quantization-guide/scripts/quantize_vit.py
Every link opens the file at its source, pinned to the revision this page describes.