Tensorflow.js tf.loadGraphModel() Function

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

The .loadGraphModel() function is used to Load a graph model given a URL to the model definition.

Syntax:

tf.loadGraphModel (modelUrl, options)

Parameters:

• modelUrl: The first tensor input that can be of type string or io.IOHandler. This parameter is the URL or an io.IOHandler that helps to load the models.
• options: The second tensor input that is optional. Options are for the HTTP request, which allows sending credentials and custom headers. The types of options are –
  • requestInit: RequestInit are for HTTP requests.
  • onProgress: OnProgress is for progress callback.
  • fetchFunc: It is a function used to override the window.fetch function.
  • strict: Strict is a loading model: whether it is an extraneous weight or missing weights should trigger an Error.
  • weightPathPrefix: Path prefix is for weight files which is by default that is calculated from the path of the model JSON file.
  • fromTFHub: It is a boolean which is whether the module or model is to be loaded from TF Hub.

Return value: It returns the Promise <tf.GraphModel>.

Example 1: In this example, we are loading MobileNetV2 from a URL and making a prediction with a zeros input.

Javascript

// Importing the tensorflow.js library
import * as tf from "@tensorflow/tfjs"

// Defining tensor input elements
const modelUrl =
'https://storage.googleapis.com/tfjs-models/savedmodel/mobilenet_v2_1.0_224/model.json';

// Calling the loadGraphModel () method
const model = await tf.loadGraphModel(modelUrl);

// Printing the zeroes
const zeros = tf.zeros([1, 224, 224, 3]);
model.predict(zeros).print();

Output:

Tensor
     [[-0.1412081, -0.5656458, 0.7578365, ..., 
     -1.0148169, -0.81284, 1.1898142],]

Example 2: In this example, we are loading MobileNetV2 from a TF Hub URL and making a prediction with a zeros input.

Javascript

// Importing the tensorflow.js library
import * as tf from "@tensorflow/tfjs"

// Defining tensor input elements
const modelUrl =
'https://tfhub.dev/google/imagenet/mobilenet_v2_140_224/classification/2';

// Calling the loadGraphModel () method
const model = await tf.loadGraphModel(
        modelUrl, {fromTFHub: true});

// Printing the zeores
const zeros = tf.zeros([1, 224, 224, 3]);
model.predict(zeros).print();

Output:

Tensor
     [[-1.0764486, 0.0097444, 1.1630495, ..., 
     -0.345558, 0.035432, 0.9112286],]

Reference: https://js.tensorflow.org/api/1.0.0/#loadGraphModel