MLproject

# ML Project:
#
# The following file specifies the project as a reproducible
# MLFlow project, which allow us to parametrize the inputs on
# consecutive runs and to track results using a tracking server.
#
# It has been created following this guide:
# https://www.mlflow.org/docs/latest/projects.html

name: madminer-workflow-ml


entry_points:

  sample:
    command: " \
      python3 {project_path}/code/sampling.py \
        --data_file {data_file} \
        --inputs_file {inputs_file} \
        --output_path {output_folder} \
        --n_samples_train {n_samples_train} \
        --n_sampling_runs {n_sampling_runs} \
        --nuisance {nuisance_flag} \
        --test_split {test_split}
      "
    parameters:
      project_path: {type: path}
      data_file: {type: path}
      inputs_file: {type: path}
      output_folder: {type: path}
      n_samples_train: {type: float, default: 1000}
      n_sampling_runs: {type: float, default: 1}
      nuisance_flag: {type: float, default: 0}
      test_split: {type: float, default: 0.5}

  train:
    command: " \
      python3 {project_path}/code/train.py \
        --alpha {alpha} \
        --batch_size {batch_size} \
        --num_epochs {num_epochs} \
        --valid_split {valid_split} \
        --inputs_file {inputs_file} \
        --samples_path {train_folder} \
        --output_path {output_folder}
      "
    parameters:
      project_path: {type: path}
      inputs_file: {type: path}
      train_folder: {type: path}
      output_folder: {type: path}
      alpha: {type: float, default: 10.0}
      batch_size: {type: float, default: 100}
      num_epochs: {type: float, default: 2}
      valid_split: {type: float, default: 0.3}

  eval:
    command: " \
      python3 {project_path}/code/evaluation.py \
        --data_file {data_file} \
        --eval_folder {eval_folder} \
        --inputs_file {inputs_file} \
        --output_path {output_folder} \
        --luminosity {luminosity} \
        --n_samples_test {n_samples_test} \
        --n_samples_train {n_samples_train} \
        --test_split {test_split}
      "
    parameters:
      project_path: {type: path}
      data_file: {type: path}
      eval_folder: {type: path}
      inputs_file: {type: path}
      output_folder: {type: path}
      luminosity: {type: float, default: 10000.0}
      n_samples_test: {type: float, default: 100}
      n_samples_train: {type: float, default: 1000}
      test_split: {type: float, default: 0.5}