electricity_elastic_net_ensemble.yaml

# @package _global_

# example hyperparameter optimization of some experiment with Optuna:
# python train.py -m hparams_search=mnist_optuna experiment=example

defaults:
  - electricity_elastic_net.yaml

model_type: "ensemble"

LOOKBACK_MIN_LENGTH: 100
MAX_LAGS: ${eval:'${LOOKBACK_MIN_LENGTH} - ${eval.kwargs.forecast_horizon}'}

# here we define Optuna hyperparameter search
# it optimizes for value returned from function with @hydra.main decorator
# docs: https://hydra.cc/docs/next/plugins/optuna_sweeper
hydra:
  sweeper:
    # name of the study to persist optimization results
    study_name: "electricity-elastic-net-ensemble"
    # define hyperparameter search space
    params:
      ++model.lags: range(2, ${MAX_LAGS}, 2) # between 2 hours and MAX_LAGS, with 1 hour increments
      ++model.output_chunk_length: choice(1, 6, 12, 24)
      ++model.model.l1_ratio: interval(0.0, 1.0)
      ++datamodule.chunk_length: range(336, 2688, 48) # between 7 and 48 days with 1 day increments
      ++ensemble.lookback_data_length: range(${LOOKBACK_MIN_LENGTH}, ${LOOKBACK_MAX_LENGTH}, 8)