NN arch implementaion
Previously, specific NNarch such as ResNetArch import tf.keras.Model.A new parent architecture called NNArch has been created and the specific architectures now inherit from it.
As of now, this class just implements a frontend block that can be added by calling the model instance. This frontend block can be composed of any number of layers. This block if created will run before any other layers in the network.
Implementation is fully backwards compatible
See discussion: #147 (closed)
assigned to @kirsebom
Looks good, all tests passing at my end.
Two minor things:
-
The new NNArch class still needs proper doc strings: https://gitlab.meridian.cs.dal.ca/public_projects/ketos/-/blob/NNArch_implementaion/ketos/neural_networks/dev_utils/nn_interface.py#L89
-
The test fixture still needs a bit of tidying (e.g. removing print statements and commented out code): https://gitlab.meridian.cs.dal.ca/public_projects/ketos/-/blob/NNArch_implementaion/ketos/tests/neural_networks/test_nn_interface.py#L59
-
enabled an automatic merge when the pipeline for f9287192 succeeds
mentioned in issue #147 (closed)
mentioned in commit fcbe5518
