Функция прогнозирования BERT - PullRequest
Новая
       45

Функция прогнозирования BERT

0 голосов
/ 24 февраля 2020 
def getPrediction(pred_sentences_A):
  #A list to map the actual labels to the predictions
  labels = ["0", "1", "2", "3", "4"]

  #Transforming the test data into BERT accepted form
  #input_examples = [run_classifier.InputExample(guid=None, text_a = x, text_b = y, label = 0) for x, y in zip(pred_sentences_A,pred_sentences_B)] 
  input_examples = [run_classifier.InputExample(guid=None, text_a = x, text_b = None, label = 0) for x in pred_sentences_A] 
  #Creating input features for Test data
  input_features = run_classifier.convert_examples_to_features(input_examples, label_list, MAX_SEQ_LENGTH, tokenizer)

  #Predicting the classes 
  predict_input_fn = run_classifier.input_fn_builder(features=input_features, seq_length=MAX_SEQ_LENGTH, is_training=False, drop_remainder=False)
  predictions = estimator.predict(input_fn=predict_input_fn, yield_single_examples=False)
  return [(sentence_A, prediction['probabilities'], prediction['labels'], labels[prediction['labels']]) for sentence_A, prediction in zip(pred_sentences_A, predictions)]

Я удивлен, увидев, что оператор return возвращает прогноз ['вероятности'], прогноз ['метки'] и так далее, где строка кода выше который присваивает предсказания = estimator.predict, а не предсказание . Тогда как получить доступ к вероятностям и меткам? Для дальнейшего использования у меня есть функция estimator.predict ниже: 

def model_fn_builder(num_labels, learning_rate, num_train_steps,
                     num_warmup_steps):
  """Returns `model_fn` closure for TPUEstimator."""
  def model_fn(features, labels, mode, params):  # pylint: disable=unused-argument
    """The `model_fn` for TPUEstimator."""

    input_ids = features["input_ids"]
    input_mask = features["input_mask"]
    segment_ids = features["segment_ids"]
    label_ids = features["label_ids"]

    is_predicting = (mode == tf.estimator.ModeKeys.PREDICT)

    # TRAIN and EVAL
    if not is_predicting:

      (loss, predicted_labels, log_probs) = create_model(
        is_predicting, input_ids, input_mask, segment_ids, label_ids, num_labels)

      train_op = bert.optimization.create_optimizer(
          loss, learning_rate, num_train_steps, num_warmup_steps, use_tpu=False)

      # Calculate evaluation metrics. 
      def metric_fn(label_ids, predicted_labels):
        accuracy = tf.metrics.accuracy(label_ids, predicted_labels)
        true_pos = tf.metrics.true_positives(
            label_ids,
            predicted_labels)
        true_neg = tf.metrics.true_negatives(
            label_ids,
            predicted_labels)   
        false_pos = tf.metrics.false_positives(
            label_ids,
            predicted_labels)  
        false_neg = tf.metrics.false_negatives(
            label_ids,
            predicted_labels)

        return {
            "eval_accuracy": accuracy,
            "true_positives": true_pos,
            "true_negatives": true_neg,
            "false_positives": false_pos,
            "false_negatives": false_neg,
            }

      eval_metrics = metric_fn(label_ids, predicted_labels)

      if mode == tf.estimator.ModeKeys.TRAIN:
        return tf.estimator.EstimatorSpec(mode=mode,
          loss=loss,
          train_op=train_op)
      else:
          return tf.estimator.EstimatorSpec(mode=mode,
            loss=loss,
            eval_metric_ops=eval_metrics)
    else:
      (predicted_labels, log_probs, output_layer) = create_model(
        is_predicting, input_ids, input_mask, segment_ids, label_ids, num_labels)
      predictions = {
          'probabilities': log_probs,
          'labels': predicted_labels,
          'pooled_output': output_layer
      }
      return tf.estimator.EstimatorSpec(mode, predictions=predictions)

  # Return the actual model function in the closure
  return model_fn
...