In the assignment I used a model with Conv1D layer followed by 2 dense layers, I get an error as below saying logits and labels must have same shape.
I avoided the error using GlobalMaxPooling layer.
I wanted to know why does GlobalMaxPooling affects the model and whether Conv1D be applied without a MaxPooling layer.
You have to check the shapes of outputs coming out of these layers, as you might know these layers change the shape of data after going through them, and layers connecting each other need to comply with shapes.
So the best way is to check it on your example by printing the shape of the transformation.
Also if predictions and labels are not of the same size then the loss cannot be computed, right!
Yeah got it, thank you for the explanation