Tensorflow.js tf.layers.gru() Function

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.

Tensorflow.js tf.layers.gru() function is used to create a RNN layer which consists of only one GRUCell and the apply method of this layer operates on a sequence of input tensors. The shape of input tensor must be atleast 2D and the first dimension must be time steps. gru is Gated Recurrent Unit.

Syntax:

tf.layers.gru(args)

Parameters:

• args: It specifies the given config object.
  1. recurrentActivation: It specifies the activation function which will be used for the recurrent step. The default value of this parameter is hard sigmoid.
  2. implementation: It specifies the implementation mode. It can be either 1 or 2. For superior performance implementation is recommended.

Return value: It returns a tf.layers.Layer

Example 1:

// Importing the tensorflow.js library 
const tf = require("@tensorflow/tfjs"); 

// Create a RNN model with gru Layer
const RNN = tf.layers.gru({units: 8, returnSequences: true});

// Create an input which will have 5 time steps
const input = tf.input({shape: [5, 10]});
const output = RNN.apply(input);

console.log(JSON.stringify(output.shape));

Output:

[null, 5, 8]

Example 2:

// Importing the tensorflow.js library 
const tf = require("@tensorflow/tfjs"); 

// Create a new model with gru Layer
const rnn = tf.layers.gru({units: 4, returnSequences: true});

// Create a 3d tensor
const x = tf.tensor3d([
    [
        [1, 2],
        [3, 4],
    ],
    [
        [5, 6],
        [7, 8],
    ],
]);

// Apply gru layer to x
const output = rnn.apply(x);

// Print output
output.print()

Output:

Reference: https://js.tensorflow.org/api/1.0.0/#layers.gru