search

timsainb / tensorflow2-generative-models

Implementations of a number of generative models in Tensorflow 2. GAN, VAE, Seq2Seq, VAEGAN, GAIA, Spectrogram Inversion. Everything is self contained in a jupyter notebook for easy export to colab.
996 stars 187 forks source link

Incorrect loss for AutoEncoder #9

Open ncoop57 opened 5 years ago

ncoop57 commented 5 years ago

In computing the testing loss for the autoencoder, you reuse the training set

# test on holdout

loss = []
    for batch, test_x in tqdm(
        zip(range(N_TRAIN_BATCHES), train_dataset), total=N_TRAIN_BATCHES # Why is this using training again?
    ):
timsainb commented 5 years ago

Thanks for pointing this out, it looks like I did this for multiple models.

ncoop57 commented 5 years ago

No problem. Thank you for this awesome set of models and straightforward code! It has greatly helped my understanding of these models, especially VAEs.