tf.data.Dataset.from_tensor_slices() Function Tensorflow.

Question

tf.data.Dataset.from_tensor_slices()

Answer 1

tf.data.Dataset.from_tensor_slices: tf.data.Dataset.from_tensor_slices() return slices of an array in object form. 

Basically when we convert a complex input pipeline into a simple input pipeline, we use tf.data API in tensorflow.

For example for text data we do preprocessing of data : removing symbols, embedding  and batching different length sequences together etc...

So, basically with help of tf.data  we can handle these complex processes and  large amount of data.

To convert a complex input pipeline into simple input pipeline, first we need data. For example 

We use tf.data.Dataset.from_tensors() or tf.data.Dataset.from_tensor_slices() to construct a Dataset from data in the memory.

Now let's see an example.

#import tensorflow  import tensorflow as tf   # use of  tf.data.Dataset.from_tensor_slices method dataset = tf.data.Dataset.from_tensor_slices([1, 2, 3, 4, 5, 6]) for elem in dataset:   print(elem.numpy())

Output

# import tensorflow  import tensorflow as tf   # use of  tf.data.Dataset.from_tensor_slices method dataset = tf.data.Dataset.from_tensor_slices([[8, 3] , [9, 2]]) print(next(iter(dataset)))

Output

tf.Tensor([8 3], shape=(2,), dtype=int32)

Note: Actually creating a simple pipeline in machine learning is always challenging then model building. we can use tf.data for data in tensorflow.