Skip to content

GitLab

Merged
Created by Oliver Kirsebom@kirsebomOwner

Methods in ResNet module for modifying momentum and dropout parameters

This merge requests adds a couple of useful methods to the resnet.py module:

  1. Method set_batch_norm_momentum in ResNetArch for modifying the momentum parameter of the batch normalization layers in the network.
  2. Method set_dropout_rate in ResNetArch for modifying the dropout rate parameter of the dropout layers in the network.
  3. Equivalent methods in ResNetBlock
  4. Possibility to specify the above parameters at initialization
  5. Added training=training in all calls to the dropout layers

The above changes have only been implemented for ResNet (2D). Same methods should be implemented for ResNet1D before merging this

Edited by Oliver Kirsebom