Use Tensorflow's Keras sub-package if the Keras package is not available.

[fabianDadada]

Hey :-) Thanks, for livelossplot!

The Tensorflow Python package includes a keras sub-package. If one only wants to use Keras with the Tensorflow backend, it is not necessary to install the Keras package, but one can use the sub-package.

[stared]

@fabianDadada A fair point, especially as Keras became (also) a part of the official TensorFlow API.

However, I still think what is the best way to incorporate that. If someone has (standalone) Keras but wants to TF, isn't it likely to cause problems? (As it will use a different Keras version than intended.)

Also, I think if there is a way to import TF/Keras only when using appropriate modules. That way it would be simple to:

• avoid loading Keras/TF when it is not necessary
• have dedicated Keras nad TFKeras variants, making the difference explicit

[stared]

Also, I asked a question on StackOverflow: https://stackoverflow.com/questions/53958762/conditional-import-in-python-when-creating-an-object-inheriting-from-it

[stared]

Regardless, the current version would need changes, as it assumes that there is Keras or TensorFlow.

[stared]

@fabianDadada If interested, feel invited to help me with #36.