Added conditional layer for extracting flattened feature vector before classification layer to the Phinet and Xinet networks.

[Sebo-the-tramp]

Description

Phinet and Xinet networks are modified to expose a conditional layer, that will return the flattened feature vector before the classification stage for when the classification head is not needed. Furthermore a new variable "features_dim" is added in the class to facilitate access to the size of the feature vector at runtime.

Also implemented the fixes from issue #85

Type of change

Please delete options that are not relevant.

• New feature (non-breaking change which adds functionality)
• This change requires a documentation update

How Has This Been Tested?

• Ran both Phinet and Xinet with solo learn
• Run image classification test on CIFAR10 both for Xinet and Phinet.

Checklist:

• My code follows the style guidelines of this project
• I have performed a self-review of my code
• My changes generate no new warnings