Shape mismatch while training

[aakashthapa22]

ValueError: Shape mismatch: The shape of labels (received (49545216,)) should equal the shape of logits except for the last dimension (received (3538944, 14)).

This occurs when I use SparseCategoricalCrossentropy while training images. However, if I use CategoricalCrossentropy there is no issue of shape mismatch but classification is not accurate.