JSONDecodeError: Expecting value when using lgb.plot_tree

[linuxTown]

Issue

Hi, issue I am experiencing is very similar to issue in #1797. I can't use lgb.plot_tree as it always crashes on parsing JSON file. I have about 90 features, from which when I print out LightGBMRegressor feature_importance, about 50 are used in prediction making.

Environment info

Operating System: Ubuntu 18.04

LightGBM version: 2.2.3 installed via pip

Error message

This is the message I receive in logs:

....
  File "/usr/lib/python3/dist-packages/simplejson/decoder.py", line 370, in decode
    obj, end = self.raw_decode(s)
  File "/usr/lib/python3/dist-packages/simplejson/decoder.py", line 400, in raw_decode
    return self.scan_once(s, idx=_w(s, idx).end())
simplejson.errors.JSONDecodeError: Expecting value: line 26 column 14 (char 2057)

Somehow after fitting LightGBM generates invalid JSON file:

Why split_gain value is nan as it is invalid JSON value?

gbm = lgb.LGBMRegressor()
gbm_fitted = gbm.fit(X_train, y_train,
                         eval_set=[(X_val, y_val)],
                         eval_metric='rmse',
                         early_stopping_rounds=20, feature_name=column_names)
... plotting features, importances...
lgb.plot_tree(gbm_fitted) # <- this is where I receive error

Interesting part is that XGBoost plot_tree with same data after training is able to plot decision tree perfectly.

What could I do to make LGBM plot this tree?