cfg.yaml

random_seed: 42

dataset_path: '../data/preprocessed.parquet'
data_cfg_path: '../data/dataset_cfg.yaml'
expert_folder_path:  '../synthetic_experts/expert_team'
destination_path_train: './train_alert'
destination_path_test: './test'
random_seed: 42

#Define which partition of the dataset should be used to fit the values of beta_0 and beta_1
#If the dataset has a timestamp column, training_set should be defined as the dates that delimit the partition
#If the dataset does not have a timestamp column, training_set should be defined as the indexes that delimit the partition
#Note - The intervals are defined as [start,end) - the last value is not included
training_set: [0,8000] 
test_set: [8000,10000]
timestamp_constraint: True

environments_train:
  batch:
    shuffle_1:
      size: 500
      seed: 42
  capacity:
    team_1:
      deferral_rate: 1
      distribution: 'homogeneous'

environments_test:
  batch:
    shuffle-1:
      size: 500
      seed: 42
  capacity:
    team_1-hom:
      deferral_rate: 0.8
      n_experts: 5
      n_experts_seed: 42
      variable_experts_per_batch: True
      distribution: 'homogeneous'