add average_output & objective fields to JSON format

[dmitryikh]

Hi!

I would like to complete JSON format with average_output (bool) and objective (string) fields.
Without these fields it is impossible to restore model behavior from JSON format.

What do you think about this?

[StrikerRUS]

Hi @dmitryikh !

Sorry for the delay and the inconvenience, but could you please rebase the branch against the latest master with the aim to make CI checks green? Thanks!

[StrikerRUS]

@guolinke Please take a look.

[dmitryikh]

@StrikerRUS , done!

[StrikerRUS]

@guolinke Can this be treated as a fix for #1700?
I mean, now we are able to get the objective via dump_model():

import numpy as np
import pandas as pd
import lightgbm as lgb

full_data = pd.DataFrame({'x1': np.random.rand(100),
                          'x2': 5 + np.random.rand(100),
                          'target': np.random.randint(0, 2, 100)})

data_train = lgb.Dataset(full_data.drop(['target'], axis=1), full_data.target)

booster = lgb.train({'objective': 'binary'}, data_train)
booster.dump_model()['objective']
>>> 'binary sigmoid:1'

...

booster = lgb.train({}, data_train)
booster.dump_model()['objective']
>>> 'regression'

[guolinke]

Yeah, good point.