etna
etna copied to clipboard
tinkoff-ai
ETNA – Time-Series Library
### 🚀 Feature Request As a newbie you can just prompt feature names and model just from general knowledge and get proper pipeline or forecast. ### Proposal ```` import os...
### 🚀 Feature Request Currently model `etna.models.SARIMAXModel` produces a lot of logs during fitting. We should give a possibility to disable it. ### Proposal Add parameter `fit_params`, that should accept...
### 🐛 Bug Report ResampleWithDistributionTransform doesn't work correctly with current behaviour of HolidaysTransform ### Expected behavior Cast columns to numerical types before resampling. It should like in `_SklearnAdapter`. ### How...
### 🚀 Feature Request In current version we have `etna.transforms.base.FutureMixin` class. We doesn't need it anymore. ### Proposal - Remove `FutureMixin` from all the classes there it is used. -...
### 🚀 Feature Request It seems like the only thing that stops us from updating torch is saving on disk. We should fix this by changing `SaveNNMixin`. Using `cloudpickle` fixes...
### 🐛 Bug Report `DeepARModel` and `TFTModel` don't work on `pytorch_lightning>=1.9.1`. Fitting fails with error: ``` AttributeError: 'tuple' object has no attribute 'items' ``` As I understand, it is connected...
### 🐛 Bug Report `MRMRFeatureSelectionTransform` fails on `fit` with `redundancy_aggregation_mode="median"`. I'm not really sure, that the problem is with `"median"` value. ### Expected behavior Works fine. ### How To Reproduce...
### 🚀 Feature Request Currently we have this condition on loading object saved under different library version: ```python if current_etna_version[0] != saved_etna_version[0] or current_etna_version[:2] < saved_etna_version[:2]: current_etna_version_str = ".".join([str(x) for...
### 🐛 Bug Report We want to distinguish between transforms ```python transform_1 = ChangePointsTrendTransform( in_column="target", change_points_model=RupturesChangePointsModel( change_points_model=Binseg(model="ar"), n_bkps=5, ), ) transform_2 = ChangePointsTrendTransform( in_column="target", change_points_model=RupturesChangePointsModel( change_points_model=Binseg(model="l1"), n_bkps=5, ) ) ```...
### 🐛 Bug Report Feature selection transforms with option `return_features=True` work incorrectly during `inverse_transform`. During `transform` it stores all the removed features inside transform and returns it back during `inverse_transform`....