Skip to content

WIP Add custom KL loss layer HLS implementation

Javier Duarte requested to merge github/fork/katyagovorkova/master into main

Created by: katyagovorkova

Adds an implementation of the KL loss layer used for CMS Anomaly detection at L1. Adds an example of usage of KL layer is in contrib/kl_layer.py and the HLS part is in hls4ml/templates/vivado/nnet_utils/nnet_distance.h. The original implementation of the KL layer is available on the AE_L1_paper branch, this PR updates the implementation for the new layer API.

Type of change

  • Documentation update
  • New feature (non-breaking change which adds functionality)

Tests

The test creates a dummy Keras model which includes the KL loss layer, converts it to an hls4ml model and synthesises it.

Test Configuration: To run the test do python contrib/kl_layer.py

Checklist

  • I have read the guidelines for contributing.
  • I have commented my code, particularly in hard-to-understand areas.
  • I have made corresponding changes to the documentation.
  • My changes generate no new warnings.
  • I have added tests that prove my fix is effective or that my feature works.

Merge request reports