mpezeshki / VAE

Example of a Variational-Autoencoder using Theano blocks

VAE

Example of a Variational-Autoencoder using Theano blocks

Dependencies

• Blocks follow the install instructions. This will install all the other dependencies for you (Theano, Fuel, etc.).

Notes

• This is a work in progress
• dropout does not work for now

Example

> python VAE.py --gamma 0.01 --batch_size 1000
    Training status:
     batch_interrupt_received: False
     epoch_interrupt_received: False
     epoch_started: False
     epochs_done: 1000
     iterations_done: 60000
     received_first_batch: True
     training_started: True
Log records from the iteration 60000:
     saved_to: vaemnist-L784,100,20-l13-g12-b10
     test_nll_bound: 105.619911194
     test_recons_term: 79.463760376
     time_read_data_this_epoch: 0.119498491287
     time_read_data_total: 118.455649614
     time_train_this_epoch: 0.715604543686
     time_train_total: 731.838293791
     train_nll_bound: 106.093231201
     train_total_gradient_norm: 32.678276062
     training_finish_requested: True
     training_finished: True