Hi @migueldeicaza, I've been struggling to setup datasets for batching with the Tensorflow's internal dataset system.
I would say the problems lie with the Iterator where it seems to get confused about the compnents of an array when receiving a TensorDataset. It is working fine with dataset generators, such as RangeDataset. Could the Mnist example be updated to show how the file-based datasets correctly implement the dataset api for batching and shuffling?
miroslav22
commented
5 years ago
Hi @migueldeicaza, I've been struggling to setup datasets for batching with the Tensorflow's internal dataset system.
I would say the problems lie with the Iterator where it seems to get confused about the compnents of an array when receiving a TensorDataset. It is working fine with dataset generators, such as RangeDataset. Could the Mnist example be updated to show how the file-based datasets correctly implement the dataset api for batching and shuffling?
And thanks of course for your work so far.