Use dictionaries for both loss and loss_weights consistently

[ronshapiro]

No description provided.

[MarkDaoust]

I was wondering how it lines up the outputs to the losses given that the outputs are defined as: outputs=[priority_pred, department_pred]

But now I see that it's using the layer names.

Should there be an example using outputs={'priority': priority_pred, 'department': department_pred}? I find that form is always clearer for the user of a multiple output model.

[ronshapiro]

Updated the description text so it's clearer where these names come from.

I don't know what's better, specifying the names in the layer or in the model.fit() call. I leave that to you.

[MarkDaoust]

Regardless of that Idea, I think this is a good change. Let's get it merged.

[MarkDaoust]

One last think, I should have mentioned on the first pass:

Could you run the script to update the generated files, and add those to the PR as well?

I believe the command is (from the scripts/ directory):

python autogen.py add_guide functional_api.py

[MarkDaoust]

Regardless of that Idea, I think this is a good change. Let's get it merged.

[fchollet]

LGTM, please replicate the fix in the md and ipynb files.

[ronshapiro]

Done.

[MarkDaoust]

Thanks.