Constructing Ragged Tensors
Here, we will learn how to create a ragged tensor with Tensorflow.
1. Using tf.ragged.constant()
We use tf.ragged.constant() to create a ragged tensor from nested Python lists. Each nested list represents a sequence of varying length. The resulting ragged tensor accommodates these variable-length sequences.
import tensorflow as tf
# Creating a ragged tensor from a nested Python list
ragged_tensor = tf.ragged.constant([[1, 2], [3, 4, 5], [6]])
print(ragged_tensor)
Output:
<tf.RaggedTensor [[1, 2], [3, 4, 5], [6]]>
RaggedTensors are multi-dimensional tensors that can have rows of different lengths. There are multiple ways to put them together.
Pairing flat values (holding all values in a flattened list) with a row-partitioning tensor that indicates how to divide those values into rows using factory classmethods like tf.RaggedTensor.from_value_rowids, tf.RaggedTensor.from_row_lengths, and tf.RaggedTensor.from_row_splits.
2. Using tf.RaggedTensor.from_value_rowids
- Values Tensor: Create a flat tensor containing all the words.
- Value Rows Tensor: This tensor tells us which row (sentence) each word belongs to. It should have the same length as the
valuestensor. Here,0represents the first sentence,2represents the second sentence, and3represents the third sentence (since it’s a single word).
import tensorflow as tf
words = [3, 1, 4, 1, 5, 9, 2]
row_ids = [0, 0, 0, 0, 2, 2, 3]
ragged_tensor = tf.RaggedTensor.from_value_rowids(values=words, value_rowids=row_ids)
print(ragged_tensor)
Output:
<tf.RaggedTensor [[3, 1, 4, 1], [], [5, 9], [2]]>
3. Using tf.RaggedTensor.from_row_lengths
row_lengths: This is constant tensor containing information about the number of elements in each row of the desired RaggedTensor. Below, [2, 2, 2, 2] specifies that we want four rows, each with two elements.
import tensorflow as tf
# Convert from row lengths
values = tf.constant([1, 2, 3, 0, 4, 0, 5, 6])
row_lengths = tf.constant([2, 2, 2, 2])
from_row_lengths_ragged = tf.RaggedTensor.from_row_lengths(values, row_lengths)
print(from_row_lengths_ragged)
Output:
<tf.RaggedTensor [[1, 2],
[3, 0],
[4, 0],
[5, 6]]>
4. Using tf.RaggedTensor.from_row_splits
row_splits: This is a constant tensor, but instead of lengths, it contains information about where each row starts in the values tensor. Below, [0, 2, 4, 6, 8] indicates the starting index for each row: 0 (first element), 2 (third element), 4 (fifth element), and 6 (seventh element).
import tensorflow as tf
# Convert from row splits
values = tf.constant([1, 2, 3, 0, 4, 0, 5, 6])
row_splits = tf.constant([0, 2, 4, 6, 8])
from_row_splits_ragged = tf.RaggedTensor.from_row_splits(values, row_splits)
print(from_row_splits_ragged)
Output:
<tf.RaggedTensor [[1, 2],
[3, 0],
[4, 0],
[5, 6]]>
Ragged tensors in TensorFlow
Ragged tensors are a fundamental data structure in TensorFlow, especially in scenarios where data doesn’t conform to fixed shapes, such as sequences of varying lengths or nested structures. In this article, we’ll understand what ragged tensors are, why they’re useful, and provide hands-on coding examples to illustrate their usage.
Table of Content
- What are Ragged Tensors?
- Why Use Ragged Tensors?
- Constructing Ragged Tensors
- Operations on Ragged Tensors
- Passing Ragged Tensors for Training