Skip to content

New feature BatchNorm

Nianxiang Fu requested to merge (removed):batch_norm into experimental

This MR includes the following changes:

  1. Added the BatchNorm feature that enables the execution of batch normalization operations.
  2. Provided one test model: batchnorm.onnx.
  3. Added support for converting the BatchNorm feature from float to int16. For inference with integer types, the standard sqrt function cannot be used; instead, bitwise_sqrt is employed to perform square root operations with integers.

Merge request reports