Tensorflow Node-js, TypeError: Only HTTP(S) protocols are supported

Question

I am trying to load a local tensorflowjs model on my Node server but am getting the error:

UnhandledPromiseRejectionWarning: TypeError: Only HTTP(S) protocols are supported.

This is how I am importing the Tensorflowjs packages:

const tf = require('@tensorflow/tfjs');
require('@tensorflow/tfjs-node');

global.fetch = require('node-fetch')

and loading the model like this:

const ssd_model_path =
    'file://models/ssd_model/model.json'

this.model = await tf.loadLayersModel(ssd_model_path);

My dependencies are as follows:

"dependencies": {
    "@tensorflow/tfjs": "^1.5.2",
    "@tensorflow/tfjs-core": "^1.2.11",
    "@tensorflow/tfjs-node": "^1.2.3",
    "electron-reload": "^1.5.0",
    "esm": "^3.2.25",
    "jimp": "^0.9.3",
    "node-fetch": "^2.6.0"
  }

Reactgular · Accepted Answer · 2020-02-09 14:22:45Z

5

I think tfjs-node is for NodeJS, and tfjs is for the web browser. You shouldn't use both at the same time, because tfjs-node has every thing as dependencies for NodeJS.

If you look at a NodeJS example, then you'll see doesn't use tfjs

https://github.com/tensorflow/tfjs-examples/tree/master/mnist-node

answered Feb 9, 2020 at 14:22

Reactgular

55.1k24 gold badges175 silver badges219 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Nachos · Accepted Answer · 2022-08-30 12:00:57Z

Not sure this exactly answers your question, but it might be slightly related. For us the issue, was that we had a library, say LIB, which invoked tf.loadLayersModel. We were using LIB in the BROWSER (browser code) and SERVER (back-end server). We basically wanted to use both tfjs and tfjs-node at the "same time", meaning we wanted to make a common library for both front and back end that used tfjs.

Note that we were only calling the tf.loadLayersModel code in LIB only through the browser.

To make the LIB --- more specifically tfjs --- compatible with the BROWSER and the SERVER, we solved it by adding the following conditional import:

// Only use tfjs-node env in node env
// Note, due to webpack? checks we
// put the string outside of the import
// to take advantage of lazy loading.
if (typeof window === 'undefined') {
    const tfjsNode = '@tensorflow/tfjs-node'
    await import(tfjsNode)
}

This allowed us to in the LIB have the load layers code

this.model = await tf.loadLayersModel('file://modelsPath/model.json');

And then call it in the SERVER to load from the file system.

Collectives™ on Stack Overflow

Tensorflow Node-js, TypeError: Only HTTP(S) protocols are supported

2 Answers 2

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related