Evaluation of a model

I have a RNN model with a pre-defined batch_size. Now I want to use the trained model to predict some data. If the number of the data is not divided by the batch_size, how should I evaluate all of them?

What do you mean by “pre-defined” batch size?
Usually the batch size is not a fixed number and can be changed during training and evaluation.
So if your last batch is smaller, it shouldn’t be a problem.