The lab says that training will take 20-30 minutes, but It’s been over an hour now and it is still training. The lab shows me as having passed with 75% but I suspect my 2 hours will time out before it finishes. (especially since it says deployment will take 20 minutes and I have barely 30 minutes left while it’s still training.
Hi Ron! Welcome to Discourse! This might be related to this problem. Kindly monitor the said thread for updates. We’ve already escalated the issue to Qwiklabs so they can check what might be wrong. We’re not yet sure if it’s just the grader or the backend code itself. Some learners have also encountered this very long training time in the past and it was resolved by simply retrying. But I advise waiting for the grader problem resolution first before doing so. I just did the lab yesterday to check the grader issue and it took around 40 minutes to train and 20 minutes to deploy. Hope this helps!
It’s more than that though. I got it to run a second attempt but correct classifications were being marked as wrong.
Hi Ron. Can you elaborate on that? As far as I know, you can only find out if it’s marked as wrong when you click the checkpoint on the Qwiklabs page. That’s what we’ve reported in the other thread but we’re still waiting for a resolution.
Or did you mean you submitted the test cloud pictures but the model is misclassifying them (i.e. on the GCP AutoML Vision UI page)? In this case, you might have re-labeled the dataset before training. This has happened before and the solution is to train the model without editing the dataset labels.
I managed to get the model deployed and the model classified correctly. When I hit the button on the lab to verify it told me “classified incorrectly or confidence under .95” even though neither of those were true.
Ah yes. That is also the issue reported in the other thread. Kindly monitor that for updates. You can click on the bell icon and set it to Watching
if you want.
Also as mentioned there, you should still get a passing grade even if the last checkpoint is currently broken. Your score should still reflect in the Grades tab as Passed. It won’t be a blocker to technically completing the course but you can surely retake it later when it comes back up. Thank you!