Disentangle graph and model config.

[dietervdb-meteo]

Is your feature request related to a problem? Please describe.

This is a minor issue, but might still be worth considering.

When an instance of AnemoiModelEncProcDec gets initiated the names of the data and hidden nodes of the graph are looked up in the graph config:

self._graph_name_data = model_config.graph.data
self._graph_name_hidden = model_config.graph.hidden

This is counter intuitive, especially when one would like to train the model with a pre-created graph stored on disk. In that case one would expect the graph part of the config to simply read:

overwrite: False

Although this will lead to the graph being loaded it will also produce an error, since as described above creation of the model needs the following lines to always be present in the graph config

data: name_of_data_nodes #typically 'data'
hidden: name_of_hidden_nodes #typically 'hidden'

Describe the solution you'd like

It would be cleaner if either
a) The names of hidden and data nodes are stored in the graph itself
b) These names are specified in the model config

Describe alternatives you've considered

No response

Additional context

No response

Organisation

RMI