IBM/pytorchpipe

# Load config defining tasks for training, validation and testing.
default_configs:
  vqa_med_2019/default_vqa_med_2019.yml,
  vqa_med_2019/frozen_pipelines/frozen_input_fusion_glove_lstm_vgg_att_is_cat.yml
  #vqa_med_2019/frozen_pipelines/frozen_question_categorization_glove_rnn_ffn.yml

c123_hyperparameters:
  # In here I am putting some of the hyperparameters from spreadsheet.

  question_preprocessing: &question_preprocessing lowercase, remove_punctuation, tokenize
  # Accepted formats: a,b,c or [a,b,c]
  # none | lowercase | remove_punctuation | tokenize | random_remove_stop_words | random_shuffle_words | all

  image_preprocessing: &image_preprocessing normalize
  # Accepted formats: a,b,c or [a,b,c]
  # none | random_affine | random_horizontal_flip | normalize | all

  # Final classifier: FFN.
  answer_classifier_hidden_sizes_val: &answer_classifier_hidden_sizes_val [500,500]

  batch_size: &batch_size 256
  preload_images: &preload_images False
  num_workers: &num_workers 4

# Training parameters:
training:
  task:
    batch_size: *batch_size
    categories: all
    export_sample_weights: ~/data/vqa-med/answers.c1_c2_c3_c4_binary_yn.weights.csv
    # Appy all preprocessing/data augmentations.
    question_preprocessing: *question_preprocessing
    image_preprocessing: *image_preprocessing 
    # Preload images.
    preload_images: *preload_images
    streams: 
      questions: tokenized_questions
  sampler:
    weights: ~/data/vqa-med/answers.c1_c2_c3_c4_binary_yn.weights.csv
  # Use four workers for loading images.
  dataloader:
    num_workers: *num_workers

  # Optimizer parameters:
  optimizer:
    type: Adam
    lr: 0.0001

  # Terminal conditions:
  terminal_conditions:
    loss_stop_threshold: 1.0e-3
    episode_limit: 10000
    epoch_limit: -1

# Validation parameters:
validation:
  task:
    batch_size: *batch_size
    categories: all
    # Appy all preprocessing/data augmentations.
    question_preprocessing: *question_preprocessing
    image_preprocessing: *image_preprocessing 
    # Preload images: false, as we will need them only once, at the end.
    preload_images: false
    streams: 
      questions: tokenized_questions
  dataloader:
    num_workers: 1


pipeline:

  ################# PIPE 6: C1 + C2 + C3 + C4 questions #################

  # Answer encoding.
  pipe6_c1234_answer_indexer:
    priority: 6.2
    type: LabelIndexer
    data_folder: ~/data/vqa-med
    word_mappings_file: answers.all.word.mappings.csv
    # Export mappings and size to globals.
    export_word_mappings_to_globals: True
    streams:
      inputs: answers
      outputs: answers_ids
    globals:
      vocabulary_size: vocabulary_size_c1234_binary_yn
      word_mappings: word_mappings_c1234_binary_yn


  # Model 4: FFN C123 answering
  pipe6_c1234_answer_classifier:
    priority: 6.3
    type: FeedForwardNetwork
    hidden: *answer_classifier_hidden_sizes_val
    dropout_rate: 0.5
    streams:
      inputs: concatenated_activations
      predictions: pipe6_c1234_predictions
    globals:
      input_size: concatenated_activations_size
      prediction_size: vocabulary_size_c1234_binary_yn

  pipe6_c1234_nllloss:
    priority: 6.4
    type: NLLLoss
    targets_dim: 1
    streams:
      predictions: pipe6_c1234_predictions
      targets: answers_ids
      loss: pipe6_c1234_loss

  pipe6_c1234_precision_recall:
    priority: 6.5
    type: PrecisionRecallStatistics
    use_word_mappings: True
    #show_class_scores: True
    #show_confusion_matrix: True
    streams:
      predictions: pipe6_c1234_predictions
      targets: answers_ids
    globals:
      word_mappings: word_mappings_c1234_binary_yn
    statistics:
      precision: pipe6_c1234_precision
      recall: pipe6_c1234_recall
      f1score: pipe6_c1234_f1score

  # C123 Predictions decoder.
  pipe6_c1234_binary_yn_prediction_decoder:
    priority: 6.6
    type: WordDecoder
    # Use the same word mappings as label indexer.
    import_word_mappings_from_globals: True
    streams:
      inputs: pipe6_c1234_predictions
      outputs: predicted_answers
    globals:
      word_mappings: word_mappings_c1234_binary_yn

  ################# PIPE 9: MERGE ANSWERS #################

  # Viewers.
  viewer:
    priority: 9.3
    type: StreamViewer
    input_streams:
      tokenized_questions,
      category_names, predicted_category_names,
      answers, predicted_answers


#: pipeline