GitLab

[trainers] fixed stuff on CNNTrainer, to be able to load model saved on both CPU/GPU on both CPU/GPU

Light cnn

See merge request !9

Added an option to return a latent embedding to the ConvAutoencoder class + unit test for this case

See merge request !8

[CNNTrainer] proper loading, initialization of pre-trained model, taking into account the case when the number of classes are different

Fine-tuning of autoencoders on multi-channel data

See merge request !7

autoencoders pretraining using RGB faces

See merge request !6