I am hoping there no problems with your model but if it they persist I could try upload your h5 file if you send it to me on private, and if there are is a problem, I could post a note for the staff to have a look on this issue!
Hi Latifa and Alex! The problem with the notebook is the use of the experimental module to access the SGD optimizer. That is not yet supported by the TF version used in the grader. Please use the SGD class under the optimizers module instead (i.e. just delete experimental). That should let the grader load your model. We’ll update the grader to accept the newer API but in the meantime, please use the older one first. We also placed a note in the notebook now to avoid this issue. Thank you for reporting!
Hi I seem to be having the same issue.
I’ve tried using both the experimental and released version of the stochastic gradient descent optimizer to train the model. With both attempts it’s still not recognised by the grader.
For reference on the colab notebook, i’m using tensorflow 2.9.
Any further suggestions would be greatly appreciated.
Hi Latifa and Alex! The problem with the notebook is the use of the experimental module to access the SGD optimizer. That is not yet supported by the TF version used in the grader. Please use the SGD class under the optimizers module instead (i.e. just delete experimental ). That should let the grader load your model. We’ll update the grader to accept the newer API but in the meantime, please use the older one first. We also placed a note in the notebook now to avoid this issue. Thank you for reporting!