difference between tf.keras.losses.SparseCategoricalCrossentropy() vs "sparse_categorical_crossentropy" as loss

Question

I'm implementing a multi class classification problem,

1. when I'm using my loss as tf.keras.losses.SparseCategoricalCrossentropy(), I'm getting very less accuracy

model.compile(optimizer="adam", loss=tf.keras.losses.SparseCategoricalCrossentropy(), metrics=["accuracy"])

2. And when I use loss as "sparse_categorical_crossentropy", I'm getting much higher accuracy

model.compile(optimizer="adam", loss="sparse_categorical_crossentropy", metrics=["accuracy"]) 

• Does this have to do with batch size?
• How to get same results as "sparse_categorical_crossentropy" using custom loss layer?

Answer 1

it's because when you use

loss=tf.keras.losses.SparseCategoricalCrossentropy() 

metrics should

"sparse_categorical_accuracy" not "accuracy"