catboost: NameError: name 'PY3' is not defined

Hi!

I installed catboost on Python 3.6.3 within a Docker container (python:3.6 image). The installation was done through PyPI. The library version is 0.6.3

I can import catboost, but when I do:

model = catboost.CatBoostRegressor()

I obtain an error:

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
<ipython-input-16-e4bd7d18c439> in <module>()
----> 1 model = catboost.CatBoostRegressor()

/usr/local/lib/python3.6/site-packages/catboost/core.py in __init__(self, iterations, learning_rate, depth, l2_leaf_reg, model_size_reg, rsm, loss_function, border_count, feature_border_type, fold_permutation_block_size, od_pval, od_wait, od_type, nan_mode, counter_calc_method, leaf_estimation_iterations, leaf_estimation_method, thread_count, random_seed, use_best_model, verbose, logging_level, metric_period, ctr_leaf_count_limit, store_all_simple_ctr, max_ctr_complexity, has_time, one_hot_max_size, random_strength, name, ignored_features, train_dir, custom_metric, eval_metric, bagging_temperature, save_snapshot, snapshot_file, fold_len_multiplier, used_ram_limit, gpu_ram_part, allow_writing_files, approx_on_full_history, boosting_type, simple_ctr, combinations_ctr, per_feature_ctr, ctr_description, task_type, device_config, devices, bootstrap_type, max_depth, n_estimators, num_boost_round, num_trees, colsample_bylevel, random_state, reg_lambda, objective, eta, max_bin, **kwargs)
   1767             objective = None
   1768 
-> 1769         if isinstance(loss_function, str) and self._is_classification_loss(loss_function):
   1770             raise CatboostError("Invalid loss_function={}: for Regressor use RMSE, MAE, Quantile, LogLinQuantile, Poisson, MAPE, R2.".format(loss_function))
   1771         params = {}

_catboost.pyx in _catboost._CatBoostBase._is_classification_loss()

_catboost.pyx in _catboost.to_binary_str()

NameError: name 'PY3' is not defined

Same happens when I specify a loss function:

model = catboost.CatBoostRegressor(loss_function='MAPE')