W4 - Assignment WebGL Context loss / crash

Course Q&A TensorFlow: Data and Deployment Specialization Browser-based Models with TensorFlow.js
1

Hello everyone,

i got stuck with the W4 assignment and have no idea how to proceed.

If i capture less than ~220 pics in total → 40 per class no error appears and prediction looks good at first glance but autograder is not happy.

When sampling more pics this error appears in the dev console

image

And training is not possible so no chance to improve the model.

The error depends on the total amount of samples made. Same if i try to make more than 220 in one class.

Can someone help me out with a hint?

Many thanks in advance!

EDIT: Tested on 3 different browsers (chrome, vivaldi, edge) and on two different machines. Stil no luck

2

with the 5th try and 45pic’s per class autograder was merciful :wink:

For me it seems that it was an issue with not enough GPU RAM (small Intel Iris in both machines try’d out).

I found that tf.setBackend(‘cpu’) would be a solution but honestly i was not patient enough to invest 20s for each picture to capture.

Sometimes you need luck… :wink: