v1.3.1
Bug fixing:
- Compulsory label requirement if the corresponding prefactor is set to non-zero. The current behavior is when then label is missing, the corresponding term does not appear in the loss function
- Optional requirement of
loss
in dipole and polarizability training
Improvement:
- Recommend consistent TensorFlow versions for python and C++ interfaces.