model.save(url) | ValueError: Cannot find any save handlers for URL

System information

• OS Platform: Mac OSX 10.14.6
• tfjs: 2.7
• tfjs-node: 2.7

Standalone code to reproduce the issue

const serverAddr = "http://localhost:5001"
model = await tf.loadLayersModel(serverAddr + '/get-model')
model.compile({
    optimizer: tf.train.adam(),
    loss: 'categoricalCrossentropy',
    metrics: ['accuracy']
  });
model.train(...)
await model.save(serverAddr + '/upload-model')    // ValueError occurs

A flask server is setup with endpoints /get-model and /upload-model. The /get-model endpoint provides an appropriate json, and so the model compile and trains properly. However, when trying to save the model to my Flask server, I receive the following error. ValueError: Cannot find any save handlers for URL 'http://localhost:5001/upload-model' at new ValueError (/Users/colinpeppler/tfplayground/node_modules/@tensorflow/tfjs-layers/dist/tf-layers.node.js:186:28).

Running the code on the browser works good when saving to the url. I also tried saving to the filesystem and that worked fine, but the HTTP request method doesn’t work with node.

[tafsiri]

A workaround in the meantime is to use https://js.tensorflow.org/api/latest/#io.http directly and pass that to tf.loadLayersModel.

[timekarma]

This is a very common issues, most of the tfjs-examples models cannot e saved, ex jena-weather, boston-housing