How to write binary cross-entropy loss funtion in LGBM?

Question

I have a binary cross-entropy implementation in Keras. I would like to implement the same one in LGBM as a custom loss. Now I understand LGBM of course has \'binary\' object