Add Quantization

Add a quantized precision to a float recipe. Tries w8a8, falls back to w8a16, or gives up. Checks both tflite and qnn_dlc and disables the pairs that fail. Hard-requires a dataset + evaluator (authors them under supervision if missing). Best run after `validate-on-device`, but does not require it.

qualcomm Updated 256 repo stars

File contents

qualcomm/ai-hub-models commit 47c6a863fa

Frequently asked questions

npx get-skill add qualcomm/add-quantization