Tensorflow.js tf.tensor4d() Function

Last Updated : 12 May, 2021

Tensorflow.js is an open-source library developed by Google for running machine learning models and deep learning neural networks in the browser or node environment.

The .tensor4d() function is used to create a new 4-dimensional tensor with the parameters namely value, shape, and datatype.

Syntax:

tf.tensor4d(value, shape, dataType)

Parameters: This function accepts the following three parameters.

value: The value of the tensor which can be nested array of numbers, or a flat array, or a TypedArray.
shape: It takes the shape of the tensor. The tensor will infer its shape from the value if it is not provided. It is an optional parameter.
datatype: It can be a ‘float32’ or ‘int32’ or ‘bool’ or ‘complex64’ or ‘string’ value. It is an optional parameter.

Return Value: It returns the tensor of the same data type. The returned tensor will always be 4-dimensional.

Note: The 4d tensor functionality can also be achieved using tf.tensor() function, but using tf.tensor4d() makes the code easily understandable and readable.

Example 1: Here, we are creating a 4d tensor and printing it. For creating a 4d tensor we are using the .tensor4d() function, and we use .print() function to print the tensor. Here, we will pass the 4d array (i.e. nested array) to the value parameter.

Javascript

// Importing the tensorflow.js library 
import * as tf from "@tensorflow/tfjs"; 
  
// Create the tensor 
let example1 = tf.tensor4d([[ 
    [[1, 3], [2, 8]], 
    [[3, 9], [4, 2]] 
]]); 
  
// Print the tensor 
example1.print() 

Output:

Tensor
    [[[[1 , 3],
       [20, 8]],

      [[3 , 9],
       [4 , 2]]]]

Example 2: Here, in this example, we are creating the tensor where we are passing the flat array and specifying the shape parameter of the tensor. We will see the usage of shape parameter here.

Javascript

// Import the tensorflow.js library 
import * as tf from "@tensorflow/tfjs"
  
// Define the value of the tensor 
const value = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]; 
  
// Specify the shape of the tensor 
const shape = [1, 2, 6, 1]; 
  
// Create the tensor 
let example2 = tf.tensor4d(value, shape); 
  
// Print the tensor 
example2.print(); 

Output:

Tensor
    [[[[1 ],
       [2 ],
       [3 ],
       [4 ],
       [5 ],
       [6 ]],

      [[7 ],
       [8 ],
       [9 ],
       [10],
       [11],
       [12]]]]

In the above example, we have created a tensor of dimensions 1 x 2 x 6 x 1.

Example 3: Here, in this example, we will create a tensor by specifying the value, shape, and datatype. We will create the tensor where all the values are of string datatype.

Javascript

// Import the tensorflow.js library 
import * as tf from "@tensorflow/tfjs"; 
  
// Define the value of the tensor 
const value = ["C", "C++", "Java", "Python", 
"PHP", "JS", "SQL", "React"]; 
  
// Specify the shape of the tensor 
const shape = [1, 2, 4, 1]; 
  
// Create the tensor 
let example3 = tf.tensor4d(value, shape); 
  
// Print the tensor 
example3.print(); 

Output:

Tensor
    [[[['C'     ],
       ['C++'   ],
       ['Java'  ],
       ['Python']],

      [['PHP'   ],
       ['JS'    ],
       ['SQL'   ],
       ['React' ]]]]

Reference: https://js.tensorflow.org/api/latest/#tensor4d