бы, чтобы сообщение пропало вообще без ответа - на будущее может пригодиться. Попробую продублировать ещё тут. "Hi! ) Happy new Year ) np.unique(pool.get_label()) returns array(['0.0', '1.0'], dtype='<U32'). But i'm feeding Pool() constructor with np.int8 at creation time. Why can it be converted to char? Pool is quantized and saved. The only param I was using at the quantisation step was 'border_count':254*2
It's able to fit, but fails at prediction step with CatBoostError: catboost/private/libs/algo/features_data_helpers.h:111: Internal CatBoost Error (contact developers for assistance): Should be IDynamicBlockIteratorPtr<TFloatValue>.
I checked, and it's np.int8 both after pool creation and after quantization. But it becomes char after reading back from disk/instatiating. Any ideas how can I fix it, please? P.S. IDynamicBlockIteratorPtr error walks away. But pool's labels are still chars, no matter what I use: np.int8, or just a list of integers. What is wrong?"
лучше сделать issue на github с репродьюсером. из текста непонятно, в какой именно момент возникает ошибка.
Обсуждают сегодня