ReferenceError in PlatformBrowser.setTimeoutCustom

[khcyt]

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): all
• Mobile device (e.g. iPhone 8, Pixel 2, Samsung Galaxy) if the issue happens on mobile device:
• TensorFlow.js installed from (npm or script link): npm
• TensorFlow.js version (use command below): 3.21
• Browser version:
• Tensorflow.js Converter Version:

Describe the current behavior

I run tensorflow in browser a separate worker thread. This is why there is an exception in PlaformBrowser.setTimeoutCustom()

platform_browser.js:53 Uncaught (in promise) ReferenceError: window is not defined
at PlatformBrowser.setTimeoutCustom (platform_browser.js:53:9)
at tryFn (util_base.js:287:13)
at util_base.js:289:9
at new Promise ()
at repeatedTry (util_base.js:274:12)
at GPGPUContext.addItemToPoll (gpgpu_context.js:398:9)
at gpgpu_context.js:375:18
at new Promise ()
at GPGPUContext.pollFence (gpgpu_context.js:374:16)
at GPGPUContext.createAndWaitForFence (gpgpu_context.js:165:21)

Describe the expected behavior

The expected behavior would be that window is not assumed to exist.
Eiter !globalThis.window or 'undefined'== typeof window

[gyagp]

@khcyt thanks for reporting the issue! Submitted a PR as above.

[rthadur]

Related PR #6923 has been merged , closing this issue.

[google-ml-butler]

Are you satisfied with the resolution of your issue?
Yes
No

[bentzibentz]

I'm using the actual v4.1.0 version, installed via github and i get still this error, cause the fix is missing:

setTimeoutCustom(functionRef, delay) {
        if (!window || !env().getBool('USE_SETTIMEOUTCUSTOM')) {
            setTimeout(functionRef, delay);
            return;
        }

Which release i have to install to get the correct version?

[gyagp]

@bentzibentz The fix was included in v4.0.0, and I just checked the github version (tfjs-core/src/platforms/platform_browser.ts). Could you please double check?

[bentzibentz]

@gyagp maybe the problem is, my installation of tsfj installs @tensorflow/tfjs-core v3.21.0
I manually install the latest version and now it is fixed, thanks for pointing me to the right direction.

[gyagp]

yes, tfjs-core v3.21.0 still has this problem and the version after it (v4.0.0) is the first one to fix it. Glad to know you're fine now.