I surely have no idea how to implement a mask in mask_exclude_positives especially for the second mask, seems I forgot to how to do so from other labs/courses.
Take a look at these calculations, it might suggest you what’s going on.
Cheers
Well, I got stuck at the mask_exclude_positives, I have no idea how to calculate it and what functions to use for.
These are the instructions given for you:
To create the mask, you need to check if the cell is diagonal by computing
tf.eye(batch_size) ==1, or if the non-diagonal cell is greater than the diagonal with(negative_zero_on_duplicate > tf.expand_dims(positive, 1)
In other words, you just need a Python “or” operator | for these two.
As you can see in the picture, the first one creates matrix with TRUE values in the diagonal.
The second, in our case, is TRUE values on the bottom row.
These two combined with Python “or” would leave only one FALSE value in the top right corner of the matrix. Which after using tf.cast() would change True/False values to 1/0 values. And this result would be the mask_exclude_positives.
Cheers
1 Like
Thank you so much! It worked!