Skip to content

Latest commit

 

History

History
12 lines (8 loc) · 439 Bytes

README.md

File metadata and controls

12 lines (8 loc) · 439 Bytes

Quantized Ops

This folder contains kernels for quantization related ops, similar to //executorch/kernels/portable/.

How to add quantized ops

  1. Add a new operator definition in quantized.yaml

  2. Implement the kernel for this operator.

  3. Add unit test in /test directory.

  4. Start to use //executorch/kernels/quantized:generated_lib in ExecuTorch or create your own generated lib if you only need a subset of the ops.