Pedro Capelastegui

I ran into this problem and I managed to fix it by installing an M1-native JVM (azul OpenJDK, .dmg version), following the instructions from this thread: https://stackoverflow.com/questions/64788005/java-jdk-for-apple-m1-chip/64881417#64881417

For reference, here are the commands used for a release: ``` # Release to test.pypi.org # From anticipy root folder rm dist/* || true python setup.py sdist bdist_wheel twine upload...

Found several different ways to implement this model. Sometimes performance is affected in non-obvious ways. Need to prepare a battery of tests to find the optimal setup.

Solution - In forecast_models.py: - create new function, _can_use_season_weekly(), based on line 1949: ```python use_season_weekly = ( # Need more than a full week (max_date_delta > pd.Timedelta(min_periods * 7, unit='d'))...

We can reproduce the problematic set of guessed parameters with an input array containing negative values: ```python forecast_models.f_init_params_decay(a_y=[1, -1, 0]) # Out: array([ 1., nan, 0., 0.]) ```

Solution: We need to check for non-negative input data when validating input for a model_decay

The following code in forecast_models would add the feature - but we would still need to test it and check that it performs well: ```python # Hourly seasonality model def...

Running the following function at the start of plot_forecast() would meet most of the requirements: ```python def df_forecast_input_to_plottable(df_input): df_result = df_input.copy() if not 'is_actuals' in df_result: df_result['is_actuals']=True if not 'date'...

This is now easier due to #198

Duplicate of #87