Wiki Talk Comentarios Predicción de toxicidad

Ver en TensorFlow.org Ejecutar en Google Colab Ver en GitHub Descargar cuaderno

En este ejemplo, consideramos la tarea de predecir si un comentario de discusión publicado en una página de discusión de Wiki contiene contenido tóxico (es decir, contiene contenido que es "grosero, irrespetuoso o irrazonable"). Utilizamos un conjunto de datos públicos publicado por el proyecto Conversation AI , que contiene más de 100.000 comentarios de la Wikipedia en inglés que son anotados por trabajadores de la multitud (consulte el documento para conocer la metodología de etiquetado).

Uno de los desafíos con este conjunto de datos es que una proporción muy pequeña de los comentarios cubre temas delicados como la sexualidad o la religión. Como tal, entrenar un modelo de red neuronal en este conjunto de datos conduce a un rendimiento dispar en los temas sensibles más pequeños. Esto puede significar que las declaraciones inocuas sobre esos temas pueden ser marcadas incorrectamente como 'tóxicas' a tasas más altas, lo que hace que el discurso sea censurado injustamente.

Al imponer restricciones durante el entrenamiento, podemos entrenar un modelo más justo que funcione de manera más equitativa en los diferentes grupos de temas.

Usaremos la biblioteca TFCO para optimizar nuestro objetivo de equidad durante el entrenamiento.

Instalación

Primero instalemos e importemos las bibliotecas relevantes. Tenga en cuenta que es posible que deba reiniciar su colab una vez después de ejecutar la primera celda debido a paquetes desactualizados en el tiempo de ejecución. Después de hacerlo, no debería haber más problemas con las importaciones.

instalaciones de pip

Tenga en cuenta que, dependiendo de cuándo ejecute la celda a continuación, es posible que reciba una advertencia sobre la versión predeterminada de TensorFlow en Colab que cambiará pronto a TensorFlow 2.X. Puede ignorar con seguridad esa advertencia, ya que este portátil fue diseñado para ser compatible con TensorFlow 1.X y 2.X.

Importar módulos

Aunque TFCO es compatible con la ejecución ávida y de gráficos, este portátil asume que la ejecución ávida está habilitada de forma predeterminada. Para asegurarse de que nada se rompa, se habilitará la ejecución ansiosa en la celda a continuación.

Habilite versiones de ejecución e impresión ávidas

Eager execution enabled by default.
TensorFlow 2.3.2
TFMA 0.26.0
FI 0.27.0.dev

Hiperparámetros

Primero, establecemos algunos hiperparámetros necesarios para el preprocesamiento de datos y el entrenamiento del modelo.

hparams = {
    "batch_size": 128,
    "cnn_filter_sizes": [128, 128, 128],
    "cnn_kernel_sizes": [5, 5, 5],
    "cnn_pooling_sizes": [5, 5, 40],
    "constraint_learning_rate": 0.01,
    "embedding_dim": 100,
    "embedding_trainable": False,
    "learning_rate": 0.005,
    "max_num_words": 10000,
    "max_sequence_length": 250
}

Cargar y preprocesar el conjunto de datos

A continuación, descargamos el conjunto de datos y lo preprocesamos. Los conjuntos de entrenamiento, prueba y validación se proporcionan como archivos CSV separados.

toxicity_data_url = ("https://raw.githubusercontent.com/conversationai/"
                     "unintended-ml-bias-analysis/master/data/")

data_train = pd.read_csv(toxicity_data_url + "wiki_train.csv")
data_test = pd.read_csv(toxicity_data_url + "wiki_test.csv")
data_vali = pd.read_csv(toxicity_data_url + "wiki_dev.csv")

data_train.head()

La columna de comment contiene los comentarios de la discusión y la columna is_toxic indica si un comentario está anotado como tóxico o no.

En lo siguiente, nosotros:

  1. Separe las etiquetas
  2. Tokenizar los comentarios de texto
  3. Identificar comentarios que contienen términos de temas delicados.

Primero, separamos las etiquetas del tren, los conjuntos de prueba y validación. Las etiquetas son todas binarias (0 o 1).

labels_train = data_train["is_toxic"].values.reshape(-1, 1) * 1.0
labels_test = data_test["is_toxic"].values.reshape(-1, 1) * 1.0
labels_vali = data_vali["is_toxic"].values.reshape(-1, 1) * 1.0

A continuación, tokenizamos los comentarios textuales utilizando el Tokenizer proporcionado por Keras . Usamos los comentarios del conjunto de entrenamiento solo para construir un vocabulario de tokens y los usamos para convertir todos los comentarios en una secuencia (acolchada) de tokens de la misma longitud.

tokenizer = text.Tokenizer(num_words=hparams["max_num_words"])
tokenizer.fit_on_texts(data_train["comment"])

def prep_text(texts, tokenizer, max_sequence_length):
    # Turns text into into padded sequences.
    text_sequences = tokenizer.texts_to_sequences(texts)
    return sequence.pad_sequences(text_sequences, maxlen=max_sequence_length)

text_train = prep_text(data_train["comment"], tokenizer, hparams["max_sequence_length"])
text_test = prep_text(data_test["comment"], tokenizer, hparams["max_sequence_length"])
text_vali = prep_text(data_vali["comment"], tokenizer, hparams["max_sequence_length"])

Finalmente, identificamos comentarios relacionados con ciertos grupos de temas sensibles. Consideramos un subconjunto de los términos de identidad proporcionados con el conjunto de datos y los agrupamos en cuatro grandes grupos de temas: sexualidad , identidad de género , religión y raza .

terms = {
    'sexuality': ['gay', 'lesbian', 'bisexual', 'homosexual', 'straight', 'heterosexual'], 
    'gender identity': ['trans', 'transgender', 'cis', 'nonbinary'],
    'religion': ['christian', 'muslim', 'jewish', 'buddhist', 'catholic', 'protestant', 'sikh', 'taoist'],
    'race': ['african', 'african american', 'black', 'white', 'european', 'hispanic', 'latino', 'latina', 
             'latinx', 'mexican', 'canadian', 'american', 'asian', 'indian', 'middle eastern', 'chinese', 
             'japanese']}

group_names = list(terms.keys())
num_groups = len(group_names)

Luego creamos matrices de pertenencia a grupos separadas para los conjuntos de preparación, prueba y validación, donde las filas corresponden a los comentarios, las columnas corresponden a los cuatro grupos sensibles y cada entrada es un booleano que indica si el comentario contiene un término del grupo de temas.

def get_groups(text):
    # Returns a boolean NumPy array of shape (n, k), where n is the number of comments, 
    # and k is the number of groups. Each entry (i, j) indicates if the i-th comment 
    # contains a term from the j-th group.
    groups = np.zeros((text.shape[0], num_groups))
    for ii in range(num_groups):
        groups[:, ii] = text.str.contains('|'.join(terms[group_names[ii]]), case=False)
    return groups

groups_train = get_groups(data_train["comment"])
groups_test = get_groups(data_test["comment"])
groups_vali = get_groups(data_vali["comment"])

Como se muestra a continuación, los cuatro grupos de temas constituyen solo una pequeña fracción del conjunto de datos general y tienen proporciones variables de comentarios tóxicos.

print("Overall label proportion = %.1f%%" % (labels_train.mean() * 100))

group_stats = []
for ii in range(num_groups):
    group_proportion = groups_train[:, ii].mean()
    group_pos_proportion = labels_train[groups_train[:, ii] == 1].mean()
    group_stats.append([group_names[ii],
                        "%.2f%%" % (group_proportion * 100), 
                        "%.1f%%" % (group_pos_proportion * 100)])
group_stats = pd.DataFrame(group_stats, 
                           columns=["Topic group", "Group proportion", "Label proportion"])
group_stats
Overall label proportion = 9.7%

Vemos que solo el 1.3% del conjunto de datos contiene comentarios relacionados con la sexualidad. Entre ellos, el 37% de los comentarios han sido anotados como tóxicos. Tenga en cuenta que esto es significativamente mayor que la proporción total de comentarios anotados como tóxicos. Esto podría deberse a que los pocos comentarios que utilizaron esos términos de identidad lo hicieron en contextos peyorativos. Como se mencionó anteriormente, esto podría hacer que nuestro modelo clasifique erróneamente los comentarios como tóxicos de manera desproporcionada cuando incluyen esos términos. Dado que esta es la preocupación, nos aseguraremos de observar la Tasa de falsos positivos cuando evaluemos el rendimiento del modelo.

Construya el modelo de predicción de toxicidad de CNN

Habiendo preparado el conjunto de datos, ahora construimos un modelo de Keras para la predicción de la toxicidad. El modelo que usamos es una red neuronal convolucional (CNN) con la misma arquitectura utilizada por el proyecto Conversation AI para su análisis de desvanecimiento. Adaptamos el código proporcionado por ellos para construir las capas del modelo.

El modelo utiliza una capa de incrustación para convertir los tokens de texto en vectores de longitud fija. Esta capa convierte la secuencia de texto de entrada en una secuencia de vectores y los pasa a través de varias capas de operaciones de convolución y agrupación, seguidas de una capa final completamente conectada.

Hacemos uso de incrustaciones de vectores de palabras GloVe previamente entrenadas, que descargamos a continuación. Esto puede tardar unos minutos en completarse.

zip_file_url = "http://nlp.stanford.edu/data/glove.6B.zip"
zip_file = urllib.request.urlopen(zip_file_url)
archive = zipfile.ZipFile(io.BytesIO(zip_file.read()))

Usamos las incrustaciones de GloVe descargadas para crear una matriz de incrustación, donde las filas contienen las incrustaciones de palabras para los tokens en el Tokenizer del Tokenizer .

embeddings_index = {}
glove_file = "glove.6B.100d.txt"

with archive.open(glove_file) as f:
    for line in f:
        values = line.split()
        word = values[0].decode("utf-8") 
        coefs = np.asarray(values[1:], dtype="float32")
        embeddings_index[word] = coefs

embedding_matrix = np.zeros((len(tokenizer.word_index) + 1, hparams["embedding_dim"]))
num_words_in_embedding = 0
for word, i in tokenizer.word_index.items():
    embedding_vector = embeddings_index.get(word)
    if embedding_vector is not None:
        num_words_in_embedding += 1
        embedding_matrix[i] = embedding_vector

Ahora estamos listos para especificar las capas de Keras . Escribimos una función para crear un nuevo modelo, que invocaremos siempre que queramos entrenar un nuevo modelo.

def create_model():
    model = keras.Sequential()

    # Embedding layer.
    embedding_layer = layers.Embedding(
        embedding_matrix.shape[0],
        embedding_matrix.shape[1],
        weights=[embedding_matrix],
        input_length=hparams["max_sequence_length"],
        trainable=hparams['embedding_trainable'])
    model.add(embedding_layer)

    # Convolution layers.
    for filter_size, kernel_size, pool_size in zip(
        hparams['cnn_filter_sizes'], hparams['cnn_kernel_sizes'],
        hparams['cnn_pooling_sizes']):

        conv_layer = layers.Conv1D(
            filter_size, kernel_size, activation='relu', padding='same')
        model.add(conv_layer)

        pooled_layer = layers.MaxPooling1D(pool_size, padding='same')
        model.add(pooled_layer)

    # Add a flatten layer, a fully-connected layer and an output layer.
    model.add(layers.Flatten())
    model.add(layers.Dense(128, activation='relu'))
    model.add(layers.Dense(1))

    return model

También definimos un método para establecer semillas aleatorias. Esto se hace para asegurar resultados reproducibles.

def set_seeds():
  np.random.seed(121212)
  tf.compat.v1.set_random_seed(212121)

Indicadores de equidad

También escribimos funciones para trazar indicadores de equidad.

def create_examples(labels, predictions, groups, group_names):
  # Returns tf.examples with given labels, predictions, and group information.  
  examples = []
  sigmoid = lambda x: 1/(1 + np.exp(-x)) 
  for ii in range(labels.shape[0]):
    example = tf.train.Example()
    example.features.feature['toxicity'].float_list.value.append(
        labels[ii])
    example.features.feature['prediction'].float_list.value.append(
        sigmoid(predictions[ii]))  # predictions need to be in [0, 1].
    for jj in range(groups.shape[1]):
      example.features.feature[group_names[jj]].bytes_list.value.append(
          b'Yes' if groups[ii, jj] else b'No')
    examples.append(example)
  return examples
def evaluate_results(labels, predictions, groups, group_names):
  # Evaluates fairness indicators for given labels, predictions and group
  # membership info.
  examples = create_examples(labels, predictions, groups, group_names)

  # Create feature map for labels, predictions and each group.
  feature_map = {
      'prediction': tf.io.FixedLenFeature([], tf.float32),
      'toxicity': tf.io.FixedLenFeature([], tf.float32),
  }
  for group in group_names:
    feature_map[group] = tf.io.FixedLenFeature([], tf.string)

  # Serialize the examples.
  serialized_examples = [e.SerializeToString() for e in examples]

  BASE_DIR = tempfile.gettempdir()
  OUTPUT_DIR = os.path.join(BASE_DIR, 'output')

  with beam.Pipeline() as pipeline:
    model_agnostic_config = agnostic_predict.ModelAgnosticConfig(
              label_keys=['toxicity'],
              prediction_keys=['prediction'],
              feature_spec=feature_map)

    slices = [tfma.slicer.SingleSliceSpec()]
    for group in group_names:
      slices.append(
          tfma.slicer.SingleSliceSpec(columns=[group]))

    extractors = [
            model_agnostic_extractor.ModelAgnosticExtractor(
                model_agnostic_config=model_agnostic_config),
            tfma.extractors.slice_key_extractor.SliceKeyExtractor(slices)
        ]

    metrics_callbacks = [
      tfma.post_export_metrics.fairness_indicators(
          thresholds=[0.5],
          target_prediction_keys=['prediction'],
          labels_key='toxicity'),
      tfma.post_export_metrics.example_count()]

    # Create a model agnostic aggregator.
    eval_shared_model = tfma.types.EvalSharedModel(
        add_metrics_callbacks=metrics_callbacks,
        construct_fn=model_agnostic_evaluate_graph.make_construct_fn(
            add_metrics_callbacks=metrics_callbacks,
            config=model_agnostic_config))

    # Run Model Agnostic Eval.
    _ = (
        pipeline
        | beam.Create(serialized_examples)
        | 'ExtractEvaluateAndWriteResults' >>
          tfma.ExtractEvaluateAndWriteResults(
              eval_shared_model=eval_shared_model,
              output_path=OUTPUT_DIR,
              extractors=extractors,
              compute_confidence_intervals=True
          )
    )

  fairness_ind_result = tfma.load_eval_result(output_path=OUTPUT_DIR)

  # Also evaluate accuracy of the model.
  accuracy = np.mean(labels == (predictions > 0.0))

  return fairness_ind_result, accuracy
def plot_fairness_indicators(eval_result, title):
  fairness_ind_result, accuracy = eval_result
  display(HTML("<center><h2>" + title + 
               " (Accuracy = %.2f%%)" % (accuracy * 100) + "</h2></center>"))
  widget_view.render_fairness_indicator(fairness_ind_result)
def plot_multi_fairness_indicators(multi_eval_results):

  multi_results = {}
  multi_accuracy = {}
  for title, (fairness_ind_result, accuracy) in multi_eval_results.items():
    multi_results[title] = fairness_ind_result
    multi_accuracy[title] = accuracy

  title_str = "<center><h2>"
  for title in multi_eval_results.keys():
      title_str+=title + " (Accuracy = %.2f%%)" % (multi_accuracy[title] * 100) + "; "
  title_str=title_str[:-2]
  title_str+="</h2></center>"
  # fairness_ind_result, accuracy = eval_result
  display(HTML(title_str))
  widget_view.render_fairness_indicator(multi_eval_results=multi_results)

Entrenar modelo sin restricciones

Para el primer modelo que entrenamos, optimizamos una simple pérdida de entropía cruzada sin restricciones.

# Set random seed for reproducible results.
set_seeds()
# Optimizer and loss.
optimizer = tf.keras.optimizers.Adam(learning_rate=hparams["learning_rate"])
loss = lambda y_true, y_pred: tf.keras.losses.binary_crossentropy(
    y_true, y_pred, from_logits=True)

# Create, compile and fit model.
model_unconstrained = create_model()
model_unconstrained.compile(optimizer=optimizer, loss=loss)

model_unconstrained.fit(
    x=text_train, y=labels_train, batch_size=hparams["batch_size"], epochs=2)
Epoch 1/2
748/748 [==============================] - 51s 69ms/step - loss: 0.1590
Epoch 2/2
748/748 [==============================] - 48s 65ms/step - loss: 0.1217
<tensorflow.python.keras.callbacks.History at 0x7f55603a1d30>

Habiendo entrenado el modelo sin restricciones, trazamos varias métricas de evaluación para el modelo en el conjunto de prueba.

scores_unconstrained_test = model_unconstrained.predict(text_test)
eval_result_unconstrained = evaluate_results(
    labels_test, scores_unconstrained_test, groups_test, group_names)
WARNING:apache_beam.runners.interactive.interactive_environment:Dependencies required for Interactive Beam PCollection visualization are not available, please use: `pip install apache-beam[interactive]` to install necessary dependencies to enable all data visualization features.
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
INFO:tensorflow:ExampleCount post export metric: could not find any of the standard keys in predictions_dict (keys were: dict_keys(['prediction']))
INFO:tensorflow:ExampleCount post export metric: could not find any of the standard keys in predictions_dict (keys were: dict_keys(['prediction']))
INFO:tensorflow:Using the first key from predictions_dict: prediction
INFO:tensorflow:Using the first key from predictions_dict: prediction
WARNING:apache_beam.io.tfrecordio:Couldn't find python-snappy so the implementation of _TFRecordUtil._masked_crc32c is not as fast as it could be.
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.6/site-packages/tensorflow_model_analysis/writers/metrics_plots_and_validations_writer.py:113: tf_record_iterator (from tensorflow.python.lib.io.tf_record) is deprecated and will be removed in a future version.
Instructions for updating:
Use eager execution and: 
`tf.data.TFRecordDataset(path)`
WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.6/site-packages/tensorflow_model_analysis/writers/metrics_plots_and_validations_writer.py:113: tf_record_iterator (from tensorflow.python.lib.io.tf_record) is deprecated and will be removed in a future version.
Instructions for updating:
Use eager execution and: 
`tf.data.TFRecordDataset(path)`

Como se explicó anteriormente, nos estamos concentrando en la tasa de falsos positivos. En su versión actual (0.1.2), los indicadores de equidad seleccionan la tasa de falsos negativos de forma predeterminada. Después de ejecutar la línea a continuación, continúe y anule la selección de false_negative_rate y seleccione false_positive_rate para ver la métrica que nos interesa.

plot_fairness_indicators(eval_result_unconstrained, "Unconstrained")

Si bien la tasa general de falsos positivos es inferior al 2%, la tasa de falsos positivos en los comentarios relacionados con la sexualidad es significativamente mayor. Esto se debe a que el grupo de sexualidad es muy pequeño y tiene una fracción desproporcionadamente mayor de comentarios anotados como tóxicos. Por lo tanto, entrenar un modelo sin restricciones da como resultado que el modelo crea que los términos relacionados con la sexualidad son un fuerte indicador de toxicidad.

Entrene con limitaciones en las tasas de falsos positivos

Para evitar grandes diferencias en las tasas de falsos positivos entre diferentes grupos, a continuación entrenamos un modelo restringiendo las tasas de falsos positivos para cada grupo para que estén dentro de un límite deseado. En este caso, optimizaremos la tasa de error del modelo sujeto a que las tasas de falsos positivos por grupo sean menores o iguales al 2% .

Sin embargo, el entrenamiento en minibatches con restricciones por grupo puede ser un desafío para este conjunto de datos, ya que los grupos que deseamos restringir son todos pequeños y es probable que los minibatches individuales contengan muy pocos ejemplos de cada grupo. Por lo tanto, los gradientes que calculamos durante el entrenamiento serán ruidosos y harán que el modelo converja muy lentamente.

Para mitigar este problema, recomendamos usar dos flujos de minibatches, con el primer flujo formado como antes a partir de todo el conjunto de entrenamiento y el segundo flujo formado únicamente a partir de los ejemplos de grupos sensibles. Calcularemos el objetivo usando minibatches del primer flujo y las restricciones por grupo usando minibatches del segundo flujo. Debido a que es probable que los lotes de la segunda transmisión contengan una mayor cantidad de ejemplos de cada grupo, esperamos que nuestras actualizaciones sean menos ruidosas.

Creamos funciones independientes, etiquetas y tensores de grupos para contener los minibatches de las dos corrientes.

# Set random seed.
set_seeds()

# Features tensors.
batch_shape = (hparams["batch_size"], hparams['max_sequence_length'])
features_tensor = tf.Variable(np.zeros(batch_shape, dtype='int32'), name='x')
features_tensor_sen = tf.Variable(np.zeros(batch_shape, dtype='int32'), name='x_sen')

# Labels tensors.
batch_shape = (hparams["batch_size"], 1)
labels_tensor = tf.Variable(np.zeros(batch_shape, dtype='float32'), name='labels')
labels_tensor_sen = tf.Variable(np.zeros(batch_shape, dtype='float32'), name='labels_sen')

# Groups tensors.
batch_shape = (hparams["batch_size"], num_groups)
groups_tensor_sen = tf.Variable(np.zeros(batch_shape, dtype='float32'), name='groups_sen')

Instalamos un nuevo modelo y calculamos predicciones para minibatches a partir de las dos transmisiones.

# Create model, and separate prediction functions for the two streams. 
# For the predictions, we use a nullary function returning a Tensor to support eager mode.
model_constrained = create_model()

def predictions():
  return model_constrained(features_tensor)

def predictions_sen():
  return model_constrained(features_tensor_sen)

Luego, establecemos un problema de optimización restringido con la tasa de error como objetivo y con restricciones en la tasa de falsos positivos por grupo.

epsilon = 0.02  # Desired false-positive rate threshold.

# Set up separate contexts for the two minibatch streams.
context = tfco.rate_context(predictions, lambda:labels_tensor)
context_sen = tfco.rate_context(predictions_sen, lambda:labels_tensor_sen)

# Compute the objective using the first stream.
objective = tfco.error_rate(context)

# Compute the constraint using the second stream.
# Subset the examples belonging to the "sexuality" group from the second stream 
# and add a constraint on the group's false positive rate.
context_sen_subset = context_sen.subset(lambda: groups_tensor_sen[:, 0] > 0)
constraint = [tfco.false_positive_rate(context_sen_subset) <= epsilon]

# Create a rate minimization problem.
problem = tfco.RateMinimizationProblem(objective, constraint)

# Set up a constrained optimizer.
optimizer = tfco.ProxyLagrangianOptimizerV2(
    optimizer=tf.keras.optimizers.Adam(learning_rate=hparams["learning_rate"]),
    num_constraints=problem.num_constraints)

# List of variables to optimize include the model weights, 
# and the trainable variables from the rate minimization problem and 
# the constrained optimizer.
var_list = (model_constrained.trainable_weights + problem.trainable_variables +
            optimizer.trainable_variables())

Estamos listos para entrenar el modelo. Mantenemos un contador separado para los dos flujos de minibatch. Cada vez que realicemos una actualización de gradiente, tendremos que copiar el contenido del minibatch del primer flujo a los tensores features_tensor y labels_tensor , y el contenido del minibatch del segundo flujo a los tensores features_tensor_sen , labels_tensor_sen y groups_tensor_sen .

# Indices of sensitive group members.
protected_group_indices = np.nonzero(groups_train.sum(axis=1))[0]

num_examples = text_train.shape[0]
num_examples_sen = protected_group_indices.shape[0]
batch_size = hparams["batch_size"]

# Number of steps needed for one epoch over the training sample.
num_steps = int(num_examples / batch_size)

start_time = time.time()

# Loop over minibatches.
for batch_index in range(num_steps):
    # Indices for current minibatch in the first stream.
    batch_indices = np.arange(
        batch_index * batch_size, (batch_index + 1) * batch_size)
    batch_indices = [ind % num_examples for ind in batch_indices]

    # Indices for current minibatch in the second stream.
    batch_indices_sen = np.arange(
        batch_index * batch_size, (batch_index + 1) * batch_size)
    batch_indices_sen = [protected_group_indices[ind % num_examples_sen]
                         for ind in batch_indices_sen]

    # Assign features, labels, groups from the minibatches to the respective tensors.
    features_tensor.assign(text_train[batch_indices, :])
    labels_tensor.assign(labels_train[batch_indices])

    features_tensor_sen.assign(text_train[batch_indices_sen, :])
    labels_tensor_sen.assign(labels_train[batch_indices_sen])
    groups_tensor_sen.assign(groups_train[batch_indices_sen, :])

    # Gradient update.
    optimizer.minimize(problem, var_list=var_list)

    # Record and print batch training stats every 10 steps.
    if (batch_index + 1) % 10 == 0 or batch_index in (0, num_steps - 1):
      hinge_loss = problem.objective()
      max_violation = max(problem.constraints())

      elapsed_time = time.time() - start_time
      sys.stdout.write(
          "\rStep %d / %d: Elapsed time = %ds, Loss = %.3f, Violation = %.3f" % 
          (batch_index + 1, num_steps, elapsed_time, hinge_loss, max_violation))
Step 747 / 747: Elapsed time = 180s, Loss = 0.068, Violation = -0.020

Habiendo entrenado el modelo restringido, trazamos varias métricas de evaluación para el modelo en el conjunto de prueba.

scores_constrained_test = model_constrained.predict(text_test)
eval_result_constrained = evaluate_results(
    labels_test, scores_constrained_test, groups_test, group_names)
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring send_type hint: <class 'NoneType'>
WARNING:apache_beam.typehints.typehints:Ignoring return_type hint: <class 'NoneType'>
INFO:tensorflow:ExampleCount post export metric: could not find any of the standard keys in predictions_dict (keys were: dict_keys(['prediction']))
INFO:tensorflow:ExampleCount post export metric: could not find any of the standard keys in predictions_dict (keys were: dict_keys(['prediction']))
INFO:tensorflow:Using the first key from predictions_dict: prediction
INFO:tensorflow:Using the first key from predictions_dict: prediction
WARNING:apache_beam.io.filebasedsink:Deleting 1 existing files in target path matching: 
WARNING:apache_beam.io.filebasedsink:Deleting 1 existing files in target path matching: 
WARNING:apache_beam.io.filebasedsink:Deleting 1 existing files in target path matching:

Como la última vez, recuerde seleccionar false_positive_rate.

plot_fairness_indicators(eval_result_constrained, "Constrained")
multi_results = {
    'constrained':eval_result_constrained,
    'unconstrained':eval_result_unconstrained,
}
plot_multi_fairness_indicators(multi_eval_results=multi_results)

Como podemos ver en los indicadores de equidad, en comparación con el modelo sin restricciones, el modelo restringido arroja tasas de falsos positivos significativamente más bajas para los comentarios relacionados con la sexualidad, y lo hace con solo una ligera caída en la precisión general.