[BUG]: Unable to train model for audio classification

[Kheertik]

I'm training 6 classes of 300 samples (5 min recording each class)

Steps to reproduce the behavior:

1. Go to record for 5 minutes
2. Click on train
3. Check console to see error

Error:

@tensorflow/speech-commands Transfer recognizer created (version 0.4.0): parameters: {"sampleRateHz":44100,"fftSize":1024,"spectrogramDurationMillis":998.4580498866213}
index.bundle.js:104  Uncaught (in promise) RangeError: Maximum call stack size exceeded
    at S (index.bundle.js:104:30520)
    at index.bundle.js:104:30955
    at new t (index.bundle.js:104:31018)
    at e.loadExamples (index.bundle.js:104:58946)
    at r.<anonymous> (index.bundle.js:5632:4033)
    at Generator.next (<anonymous>)
    at a (index.bundle.js:5598:3279)
    

Expected behavior: The model should train as it should be.

Specs:

• OS: Windows 11 64 bits
• Browser: Chrome (extensions: Adblocker, noscript(every script enabled), Decentralysed), Firefox (no extension)
• Version 22H

[Miraii133]

I have the exact same problem when training a model for audio classification. I have 20 classes, and 350 sample for each class. I get that error when I reach a certain number of datasets.

Tried in different browsers: Edge, Google Chrome (fresh install), Opera (fresh install)

Any idea how to fix this?