Hi all, I can’t reach the minimum accuracy. If I change the loss function and the number of epoch I can reach a 0.92; but the assignment don’t admit this change. What I can do?
2 Likes
Hi, I have the same problem, there is no chance to increase the accuracy without modifying the training parameters, which are constrained.
1 Like
my problem was due to a definition of the model outputs
my model summary looks exactly the same as the requested
1 Like
It is very disappointed at this specialization and at deeplearning.ai. I have made a very simple question which prevents me from continue my specialization and nobody from deeplearning.ai answers. I have posted the problem to coursera and to deeplearning.ai, besides to this forum and there is no support at all. I feel that deeplearning.ai just hangs the online courses and forget completely about the participants. This would be ok for a youtube training, but it is not what I expected from a paid specialization.
1 Like
Hi there sorry for the delay, I can understand your concern.
Please share me the code in dm so that I can debug the issue.
1 Like
Hi @anscipione, @GLozano, did you try running the training a few times? Sometimes, this can surface because of how random params are being initialized.
1 Like
i’ve resolved changing conv2Trans to conv2d
2 Likes
Thank you for letting me know. I’ve marked your comment as a solution; hopefully, it will help someone else too until this is properly fixed.
1 Like
Hi all - opening up the thread again. Using conv2d does not allow to reach the accuracy needed with the model specified. I am stuck around 82% and I have retrained the model many many times to see if randomness of the initial weights is the issue. But no luck.
1 Like
Hi Sylvain and welcome to the community! Let’s continue the discussion on the new topic you opened. This might be a separate issue. Thanks!