I’ve got a model with a learning rate that produces MSE @ 5.27, MAE @ 1.79, both locally, and in coursera-hosted notebook, but every time I submit the .ipynb and the .h5, I get a 0%.
Failed test case: Mean Absolute Error is too big.
Expected:
a value less than 2,
but got:
5.25.
Failed test case: Mean Squared Error is too big.
Expected:
a value less than 6,
but got:
39.76.
I suspect it has something to do with the tar.gz, but there doesn’t seem to be a step that adds my .h5 to the tar? Perhaps I deleted a cell…
I’ve tried a billion times to submit this thing lol.
I even reset the system back to your checkpoint.
My latest error is:
There was a problem loading your model. Details:
No file or directory found at ./submission/saved_model/my_model
I don’t understand why it’s looking for a ./submission file path when its not even referenced in your .ipynb… Please, please just reach out and ask me for the .h5 and the .ipynb so I can submit it and get this course over with.
I finally got it taken care out.
Pro tip: if you’re going to offer multiple avenues for executing the homeworks (ie clone via github, google colab, coursera jupyter), (1) keep them all up to date with each other, and (2) send out notifications when changes are made, for the github family.
The problem was that here’s your github hosted “make-the-submission-payload”:
# Save your model into a HDF5 file
model.save("my_model.h5")
# Compress the directory using tar
! tar -czvf saved_model.tar.gz saved_model/
And here’s the one on the coursera notebook.
# Save your model in the SavedModel format
model.save('saved_model/my_model')
# Compress the directory using tar
! tar -czvf saved_model.tar.gz saved_model/
Was a blast spending the majority of this homework assignment debugging this problem rather than on building machine learning models.
Thanks.
Glad to see you’ve figured things out.
There’s only one link for the assignment which points to the coursera lab workspace. Could you please explain the reference to a public repo for this assignment?
Here’s the screenshot of the assignment section.
