name: Keras train classifier from csv
description: |-
  Trains classifier model using Keras.

      Annotations:
          author: Alexey Volkov <alexey.volkov@ark-kun.com>
inputs:
- {name: training_features, type: CSV}
- {name: training_labels, type: CSV}
- {name: network_json, type: KerasModelJson}
- {name: loss_name, type: String, default: categorical_crossentropy, optional: true}
- {name: num_classes, type: Integer, optional: true}
- {name: optimizer, type: String, default: rmsprop, optional: true}
- {name: optimizer_config, type: JsonObject, optional: true}
- {name: learning_rate, type: Float, default: '0.01', optional: true}
- {name: num_epochs, type: Integer, default: '100', optional: true}
- {name: batch_size, type: Integer, default: '32', optional: true}
- {name: metrics, type: JsonArray, default: '["accuracy"]', optional: true}
- {name: random_seed, type: Integer, default: '0', optional: true}
outputs:
- {name: model, type: KerasModelHdf5}
- {name: final_loss, type: Float}
- {name: final_metrics, type: JsonObject}
- {name: metrics_history, type: JsonObject}
implementation:
  container:
    image: tensorflow/tensorflow:2.2.0
    command:
    - sh
    - -c
    - (PIP_DISABLE_PIP_VERSION_CHECK=1 python3 -m pip install --quiet --no-warn-script-location
      'keras==2.3.1' 'pandas==1.0.5' || PIP_DISABLE_PIP_VERSION_CHECK=1 python3 -m
      pip install --quiet --no-warn-script-location 'keras==2.3.1' 'pandas==1.0.5'
      --user) && "$0" "$@"
    - python3
    - -u
    - -c
    - |
      def _make_parent_dirs_and_return_path(file_path: str):
          import os
          os.makedirs(os.path.dirname(file_path), exist_ok=True)
          return file_path

      def keras_train_classifier_from_csv(
          training_features_path,
          training_labels_path,
          network_json_path,
          model_path,
          loss_name = 'categorical_crossentropy',
          num_classes = None,
          optimizer = 'rmsprop',
          optimizer_config = None,
          learning_rate = 0.01,
          num_epochs = 100,
          batch_size = 32,
          metrics = ['accuracy'],
          random_seed = 0,
      ):
          '''Trains classifier model using Keras.

          Annotations:
              author: Alexey Volkov <alexey.volkov@ark-kun.com>
          '''
          from pathlib import Path

          import keras
          import numpy
          import pandas
          import tensorflow

          tensorflow.random.set_seed(random_seed)
          numpy.random.seed(random_seed)

          training_features_df = pandas.read_csv(training_features_path)
          training_labels_df = pandas.read_csv(training_labels_path)

          x_train = training_features_df.to_numpy()
          y_train_labels = training_labels_df.to_numpy()
          print('Training features shape:', x_train.shape)
          print('Numer of training samples:', x_train.shape[0])

          # Convert class vectors to binary class matrices.
          y_train_one_hot = keras.utils.to_categorical(y_train_labels, num_classes)

          model_json_str = Path(network_json_path).read_text()
          model = keras.models.model_from_json(model_json_str)

          model.add(keras.layers.Activation('softmax'))

          # Initializing the optimizer
          optimizer_config = optimizer_config or {}
          optimizer_config['learning_rate'] = learning_rate
          optimizer = keras.optimizers.deserialize({
              'class_name': optimizer,
              'config': optimizer_config,
          })

          model.compile(
              loss=loss_name,
              optimizer=optimizer,
              metrics=metrics,
          )

          history = model.fit(
              x_train,
              y_train_one_hot,
              batch_size=batch_size,
              epochs=num_epochs,
              shuffle=True
          )

          model.save(model_path)

          metrics_history = {name: [float(value) for value in values] for name, values in history.history.items()}
          final_metrics = {name: values[-1] for name, values in metrics_history.items()}
          final_loss = final_metrics['loss']
          return (final_loss, final_metrics, metrics_history)

      import json
      def _serialize_float(float_value: float) -> str:
          if isinstance(float_value, str):
              return float_value
          if not isinstance(float_value, (float, int)):
              raise TypeError('Value "{}" has type "{}" instead of float.'.format(str(float_value), str(type(float_value))))
          return str(float_value)

      def _serialize_json(obj) -> str:
          if isinstance(obj, str):
              return obj
          import json
          def default_serializer(obj):
              if hasattr(obj, 'to_struct'):
                  return obj.to_struct()
              else:
                  raise TypeError("Object of type '%s' is not JSON serializable and does not have .to_struct() method." % obj.__class__.__name__)
          return json.dumps(obj, default=default_serializer, sort_keys=True)

      import argparse
      _parser = argparse.ArgumentParser(prog='Keras train classifier from csv', description='Trains classifier model using Keras.\n\n    Annotations:\n        author: Alexey Volkov <alexey.volkov@ark-kun.com>')
      _parser.add_argument("--training-features", dest="training_features_path", type=str, required=True, default=argparse.SUPPRESS)
      _parser.add_argument("--training-labels", dest="training_labels_path", type=str, required=True, default=argparse.SUPPRESS)
      _parser.add_argument("--network-json", dest="network_json_path", type=str, required=True, default=argparse.SUPPRESS)
      _parser.add_argument("--loss-name", dest="loss_name", type=str, required=False, default=argparse.SUPPRESS)
      _parser.add_argument("--num-classes", dest="num_classes", type=int, required=False, default=argparse.SUPPRESS)
      _parser.add_argument("--optimizer", dest="optimizer", type=str, required=False, default=argparse.SUPPRESS)
      _parser.add_argument("--optimizer-config", dest="optimizer_config", type=json.loads, required=False, default=argparse.SUPPRESS)
      _parser.add_argument("--learning-rate", dest="learning_rate", type=float, required=False, default=argparse.SUPPRESS)
      _parser.add_argument("--num-epochs", dest="num_epochs", type=int, required=False, default=argparse.SUPPRESS)
      _parser.add_argument("--batch-size", dest="batch_size", type=int, required=False, default=argparse.SUPPRESS)
      _parser.add_argument("--metrics", dest="metrics", type=json.loads, required=False, default=argparse.SUPPRESS)
      _parser.add_argument("--random-seed", dest="random_seed", type=int, required=False, default=argparse.SUPPRESS)
      _parser.add_argument("--model", dest="model_path", type=_make_parent_dirs_and_return_path, required=True, default=argparse.SUPPRESS)
      _parser.add_argument("----output-paths", dest="_output_paths", type=str, nargs=3)
      _parsed_args = vars(_parser.parse_args())
      _output_files = _parsed_args.pop("_output_paths", [])

      _outputs = keras_train_classifier_from_csv(**_parsed_args)

      _output_serializers = [
          _serialize_float,
          _serialize_json,
          _serialize_json,

      ]

      import os
      for idx, output_file in enumerate(_output_files):
          try:
              os.makedirs(os.path.dirname(output_file))
          except OSError:
              pass
          with open(output_file, 'w') as f:
              f.write(_output_serializers[idx](_outputs[idx]))
    args:
    - --training-features
    - {inputPath: training_features}
    - --training-labels
    - {inputPath: training_labels}
    - --network-json
    - {inputPath: network_json}
    - if:
        cond: {isPresent: loss_name}
        then:
        - --loss-name
        - {inputValue: loss_name}
    - if:
        cond: {isPresent: num_classes}
        then:
        - --num-classes
        - {inputValue: num_classes}
    - if:
        cond: {isPresent: optimizer}
        then:
        - --optimizer
        - {inputValue: optimizer}
    - if:
        cond: {isPresent: optimizer_config}
        then:
        - --optimizer-config
        - {inputValue: optimizer_config}
    - if:
        cond: {isPresent: learning_rate}
        then:
        - --learning-rate
        - {inputValue: learning_rate}
    - if:
        cond: {isPresent: num_epochs}
        then:
        - --num-epochs
        - {inputValue: num_epochs}
    - if:
        cond: {isPresent: batch_size}
        then:
        - --batch-size
        - {inputValue: batch_size}
    - if:
        cond: {isPresent: metrics}
        then:
        - --metrics
        - {inputValue: metrics}
    - if:
        cond: {isPresent: random_seed}
        then:
        - --random-seed
        - {inputValue: random_seed}
    - --model
    - {outputPath: model}
    - '----output-paths'
    - {outputPath: final_loss}
    - {outputPath: final_metrics}
    - {outputPath: metrics_history}