Eval callback

[Mohammadamin-Barekatain]

Evaluate the single-task or multi-task model after every epoch via the given metric. Evaluation is done for each individual task and the average is also reported. Evaluation results are added to logs so that they can be used with other callbacks such as TensorBoard, EarlyStopping, ModelCheckpoint when this callback is added to callback list earlier.

[Mohammadamin-Barekatain]

Thanks for the feedback. I've updated the requested changes. I didn't do task_names=None - optional list of task_names. If None, incremental integer is used because we cannot infer the number of inputs if we don't have the tasks names.

[Avsecz]

looks great! Thanks!