How to change/insert the shape of/into BatchDataset?

[Jeff Yang]

I'm replicating deep voice 3 paper from a github repo with TF 2.0.

[https://github.com/Kyubyong/deepvoice3]

I have done all preprocessing.

I have also created from tensor (shape is (13066,)) to TensorSliceDataset (shape is (None,)), mapped some functions and already batched it.

Now, I have BatchDataset which shape is (None, None).

**I have no idea how to change that (None, None) shape into desire shape (16, 180)?**

The repo is implemented with TF 1.3.

I am using TF 2.0.

When the old code is tf.train.batch and I head to the tensorflow website and found to use tf.data.Dataset in TF 2.0.

But, it doesn't have an option to shape the dataset.

Below is TF 1.3 code.

```

# TF old version

texts = tf.train.batch([text], shapes=[(hp.Tx,)],

                               num_threads=32,

                               batch_size=hp.batch_size,

                               capacity=hp.batch_size*32,   

                               dynamic_pad=False) # (16, 180)

# TF 2.0

texts = texts.batch(hp.batch_size) # (None, None)

```

The shape of BatchDataset before applying shape is (None, None).

The shape of BatchDataset after applying shape should be (16, 180).

I have asked in StackOverflow but no one answered my question.

Thank you!