loadFrozenModel does not work in node

[DavidGOrtega]

TensorFlow.js version

tfjs 0.13.1
tfjs-node 0.1.17
tfjs-converter 0.6.2

Browser version

node 10.5.0 on Windows 10

Describe the problem or feature request

Promise { <pending> }
(node:21296) UnhandledPromiseRejectionWarning: Error: browserHTTPRequest is not supported outside the web browser without a fetch polyfill.
    at new BrowserHTTPRequest (C:\nodetfjs\node_modules\@tensorflow\tfjs-core\dist\io\browser_http.js:46:19)
    at Object.browserHTTPRequest (C:\nodetfjs\node_modules\@tensorflow\tfjs-core\dist\io\browser_http.js:247:12)
    at FrozenModel.findIOHandler (C:\nodetfjs\node_modules\@tensorflow\tfjs-converter\dist\src\executor\frozen_model.js:99:38)
    at FrozenModel.<anonymous> (C:\nodetfjs\node_modules\@tensorflow\tfjs-converter\dist\src\executor\frozen_model.js:114:30)
    at step (C:\nodetfjs\node_modules\@tensorflow\tfjs-converter\dist\src\executor\frozen_model.js:32:23)
    at Object.next (C:\nodetfjs\node_modules\@tensorflow\tfjs-converter\dist\src\executor\frozen_model.js:13:53)
    at C:\nodetfjs\node_modules\@tensorflow\tfjs-converter\dist\src\executor\frozen_model.js:7:71
    at new Promise (<anonymous>)
    at __awaiter (C:\nodetfjs\node_modules\@tensorflow\tfjs-converter\dist\src\executor\frozen_model.js:3:12)
    at FrozenModel.load (C:\nodetfjs\node_modules\@tensorflow\tfjs-converter\dist\src\executor\frozen_model.js:109:16)

Code to reproduce the bug / link to feature request

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

const URI           = 'file://C:/pix2pix/models/facades_A2B/tfjs/';
const MODEL_URL     = URI + 'tensorflowjs_model.pb';
const WEIGHTS_URL   = URI + 'weights_manifest.json';
const model         = tfcv.loadFrozenModel(MODEL_URL, WEIGHTS_URL);

you can try also with same results

const model         = tf.loadFrozenModel(MODEL_URL, WEIGHTS_URL);

[caisq]

Did you try global.fetch = require('node-fetch') (with dependency on node-fetch)?

[DavidGOrtega]

Hi @caisq . Thanks for the reply!
Yeah that one was one of my initial try outs.
With the polyfill I can load from HTTP but not local models.

Only HTTP(S) protocols are supported

I 'm looking at the code. LoadfrozenModel is just

const model = new tf.FrozenModel(MODEL_URL, WEIGHTS_URL, {});
model.load()

However it depends on a handler that is going ALWAYS to be empty on node

const handlers = tfc.io.getLoadHandlers(path);
if (handlers.length === 0) {
        handlers.push(tfc.io.browserHTTPRequest(path, this.requestOption));
         }...
         
this.handler = this.handlers[0];

My options right now are
1 - Write my frozenModel that loads the model with it's own IOHandler
2 - Write just the fileLoader and polyfill tfc.io.getLoadHandlers to return my IOHandler
3 - Try to find a fetch polyfill that loads file://

What would you do?
;)

[caisq]

Sorry I overlooked an important problem in your code before. Please import tfjs-node after importing tfjs, i.e.:

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

[DavidGOrtega]

@caisq I do it intentionally.

If I do it the other way around it does not fix the file:// issue and arises a nasty error:

Registration of backend tensorflow failed
TypeError: backend.setDataMover is not a function

cpu backend was already registered. Reusing existing backend
2018-10-04 18:22:19.784559: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX AVX2
Registration of backend tensorflow failed
TypeError: backend.setDataMover is not a function
    at Environment.registerBackend (C:\\nodetfjs\node_modules\@tensorflow\tfjs-core\dist\environment.js:186:21)
    at Object.<anonymous> (C:\\nodetfjs\node_modules\@tensorflow\tfjs-node\dist\index.js:9:9)
    at Module._compile (internal/modules/cjs/loader.js:702:30)
    at Object.Module._extensions..js (internal/modules/cjs/loader.js:713:10)
    at Module.load (internal/modules/cjs/loader.js:612:32)
    at tryModuleLoad (internal/modules/cjs/loader.js:551:12)
    at Function.Module._load (internal/modules/cjs/loader.js:543:3)
    at Module.require (internal/modules/cjs/loader.js:650:17)
    at require (internal/modules/cjs/helpers.js:20:18)
    at Object.<anonymous> (C:\\nodetfjs\test2.js:5:1)
(node:28416) UnhandledPromiseRejectionWarning: Error: file://C:/models/pix2pix/models/facades_A2B/tfjs/tensorflowjs_model.pb not found. TypeError: Only HTTP(S) protocols are supported
    at BrowserHTTPRequest.<anonymous> (C:\\nodetfjs\node_modules\@tensorflow\tfjs-core\dist\io\browser_http.js:124:31)
    at step (C:\\nodetfjs\node_modules\@tensorflow\tfjs-core\dist\io\browser_http.js:32:23)
    at Object.throw (C:\\nodetfjs\node_modules\@tensorflow\tfjs-core\dist\io\browser_http.js:13:53)
    at rejected (C:\\nodetfjs\node_modules\@tensorflow\tfjs-core\dist\io\browser_http.js:5:65)
    at process._tickCallback (internal/process/next_tick.js:68:7)
    at Function.Module.runMain (internal/modules/cjs/loader.js:746:11)
    at startup (internal/bootstrap/node.js:240:19)
    at bootstrapNodeJSCore (internal/bootstrap/node.js:564:3)
(node:28416) UnhandledPromiseRejectionWarning: Unhandled promise rejection. This error originated either by throwing inside of an async function without a catch block, or by rejecting a promise which was not handled with .catch(). (rejection id: 3)
(node:28416) [DEP0018] DeprecationWarning: Unhandled promise rejections are deprecated. In the future, promise rejections that are not handled will terminate the Node.js process with a non-zero exit code.
(node:28416) UnhandledPromiseRejectionWarning: TypeError: Only HTTP(S) protocols are supported
    at getNodeRequestOptions (C:\\nodetfjs\node_modules\node-fetch\lib\index.js:1261:9)
    at C:\\nodetfjs\node_modules\node-fetch\lib\index.js:1322:19
    at new Promise (<anonymous>)
    at fetch (C:\\nodetfjs\node_modules\node-fetch\lib\index.js:1319:9)
    at BrowserHTTPRequest.<anonymous> (C:\\nodetfjs\node_modules\@tensorflow\tfjs-core\dist\io\browser_http.js:137:36)
    at step (C:\\nodetfjs\node_modules\@tensorflow\tfjs-core\dist\io\browser_http.js:32:23)
    at Object.next (C:\\nodetfjs\node_modules\@tensorflow\tfjs-core\dist\io\browser_http.js:13:53)
    at C:\\nodetfjs\node_modules\@tensorflow\tfjs-core\dist\io\browser_http.js:7:71
    at new Promise (<anonymous>)
    at __awaiter (C:\\nodetfjs\node_modules\@tensorflow\tfjs-core\dist\io\browser_http.js:3:12)
(node:28416) UnhandledPromiseRejectionWarning: Unhandled promise rejection. This error originated either by throwing inside of an async function without a catch block, or by rejecting a promise which was not handled with .catch(). (rejection id: 4)

[caisq]

I see. I think you can remove @tensorflow/tfjs as a dependency from your package.json and keep @tensorflow/tfjs-node. You are running into these issues due to multiple versions of @tensorflow/tfjs being pulled in.

[DavidGOrtega]

I can't!

"dependencies": {
    "@tensorflow/tfjs-node": "^0.1.17",
    "node-fetch": "^2.2.0",
  },

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

Error: Cannot find module '@tensorflow/tfjs'
    at Function.Module._resolveFilename (internal/modules/cjs/loader.js:594:15)
    at Function.Module._load (internal/modules/cjs/loader.js:520:25)
    at Module.require (internal/modules/cjs/loader.js:650:17)
    at require (internal/modules/cjs/helpers.js:20:18)

[nkreeger]

Try deleting the node_modules directory and retry npm install or yarn

[DavidGOrtega]

@nkreeger thanks for the reply. I tried it and documented it #739
I did swept the node-modules completely and just installed the @tensorflow/tfjs-node however did not worked for me. I needed to install the tfjs module.

[nkreeger]

@DavidGOrtega I think you have an old version of tfjs-node. Please try installing 0.1.18.

[DavidGOrtega]

@nkreeger thanks, that solved the problem of issue #739 ! I'm impressed how fast new versions come out! mine was a couple of days old?

[jtoy]

nice, I ran into same thing and found this thread. think we can close this thread.

[rthadur]

Automatically closing due to lack of recent activity. Please update the issue when new information becomes available, and we will reopen the issue. Thanks!