Add suport for keras.Sequential Callbacks #159
Description
Add support for callbacks in the mlprimitives.adapters.keras.Sequential class.
The implementation should be as follows:
- Add a new fixed hyperparameter called
callbacks, which will contain a list. - Each element in the list is a subdocument containing:
class: name of the callback classargs: subdocument with the callback class arguments.
The args argument should be optional, and when given, it should be able to contain only the arguments that are different than the default ones.
For example, adding EearlyStoppìng with the default arguments would look like this (note how args is not given):
"callbacks": [
{
"class": "keras.callbacks.EarlyStopping"
}
]
In case some argument was different than the default value, the args subdocument would be added:
"callbacks": [
{
"class": "keras.callbacks.EarlyStopping",
"args": {
"patience": 2,
"mode": "min",
}
}
]