C3_A2_Grader_Inconsistency

Course Q&A TensorFlow Developer Professional Natural Language Processing in TensorFlow
1

To the team,

After completing the assignment and solving the issues with grader, I have found some small inconsistencies in just the fit_encoder_labels part of the assignment. I am attaching a redacted version of the graded code cell to explain my point. No user code or solution is being shown.

  1. The two sentences underlined in red contradict each other. The instructions before the cell state that the layer should be fit to all labels, but in the comments in the cell it states that it should be applied to just the train labels.
  2. This directly leads to another inconsistency when initializing the StringLookUp layer from Tensorflow. If the layer is fit to all labels, then there is no need to call additional parameters other than that mentioned in the instructions. Conversely, should a student fit to just the training labels, then while initializing the layer in the previous line an additional parameter needs to be set.
  3. Submitting the assignment in one of two ways mentioned above gives a full points in the grader output.

This is what I found after submitting the assignment 4 separate times with 4 minutely different codes.

Regards

Prajwal

Screenshot 2025-03-25 163822
Screenshot 2025-03-25 1638221512×610 25.9 KB

2

the underlined redline you are mentioning in the instruction is for concatenating the assigned training and validations labels.

and the second underline red line, use the vectorizer to the labels making sure to avoid oov labels.

So the steps are clearly mentioned and as per I saw your codes you have followed the steps except the first time where you used dataset.zip instead of the concatenate as per instructions given.

I don’t know your 4 times submission with what codes, but the one I saw you had code issue in train_val, fit_label_encoder and preprocess data.

being said that even after correction if you had encountered failed submission, always make sure to save your work before submission or in some case clearout the kernel output and restart the kernel, re-run the codes and submit. this should surely make a successful submission.