The batch_size parameter currently resides in both, the dataset and the model configuration.
This can lead to spurious errors, e.g. when evaluating the Harris WGAN model, the shape of the noise-tensor is not proper.
The batch_size should probably be defined in the dataset only.
The
batch_sizeparameter currently resides in both, the dataset and the model configuration.This can lead to spurious errors, e.g. when evaluating the Harris WGAN model, the shape of the noise-tensor is not proper.
The
batch_sizeshould probably be defined in the dataset only.