The initializers in Tensorflow.js are used to initialize the starting values of kernel, weights, and biases. The tf.initializers.constant() is initializer function inherited from Initializer base class. This function is used to generate the values initialized to some constant. In this post, we are going to know about tf.initializers.constant() function in Tensorflow.js.
Syntax:
tf.initializers.constant(args)
Parameters: The args object contains the following props.
- values: The value for each element in the variable.
Return Value: It returns the tf.initializers.Initializer.
Example 1: In this example, we are going to see standalone use of tf.initializers.constant() function.
Javascript
// Importing the tensorflow.js libraryconst tf = require("@tensorflow/tfjs")// Use tf.initializers.constant() functionvar initializer = tf.initializers.constant({ value: 7, })// Print the value of constantconsole.log(initializer); |
Output:
Constant { value: 7 }
Example 2: In this example, we are going to use the constant() function in model creation to initialize the kernel.
Javascript
// Importing the tensorflow.js libraryconst tf = require("@tensorflow/tfjs")// Using tf.initializers.constant() functionvar initializer = tf.initializers.constant({ value: 7, })// Create modelconst model = tf.sequential();// Add layer and initialize the kernelmodel.add(tf.layers.dense({ units: 3, activation: 'softmax', kernelInitializer: initializer, inputShape: [2]}));// Print the summarymodel.summary(); |
Output:
Layer (type) Output shape Param # ================================================================= dense_Dense1 (Dense) [null,3] 9 ================================================================= Total params: 9 Trainable params: 9 Non-trainable params: 0 _________________________________________________________________
Reference: https://js.tensorflow.org/api/latest/#initializers.constant
