I did tf.reduce_sum
I did tf.transpose
I did from_logits = True
But I am still getting huge difference in the results. Can anyone help me?
1 Like
Did you use tf.keras.losses.categorical_crossentropy(…) ?
1 Like
In addition to using the correct loss function as Tom says, you also need to make sure that you give the positional arguments in the expected order to the aforementioned function. That’s item 4) on the “checklist” for this function. Please make sure you read the documentation page for the loss function. They gave you the link in the instructions.
1 Like
Mistake was in positions.
I was putting logits before labels, now it is resolved.
Thanks a lot.
1 Like