Question

我尝试了以下代码来重用已保存的BERT模型。

def serving_input_receiver_fn():
  feature_spec = {
    "input_ids" : tf.FixedLenFeature([MAX_SEQ_LENGTH], tf.int64),
    "input_mask" : tf.FixedLenFeature([MAX_SEQ_LENGTH], tf.int64),
    "segment_ids" : tf.FixedLenFeature([MAX_SEQ_LENGTH], tf.int64),
    "label_ids" :  tf.FixedLenFeature([], tf.int64)
  }
  serialized_tf_example = tf.placeholder(dtype=tf.string,                                         
                                     shape=[None],
                                     name='input_example_tensor')
  print(serialized_tf_example, "serialized_tf_example")
  print(serialized_tf_example.shape, "Shape")
  receiver_tensors = {'example': serialized_tf_example}
  print(receiver_tensors, "receiver_tensors")
  features = tf.parse_example(serialized_tf_example, feature_spec)
  return tf.estimator.export.ServingInputReceiver(features, receiver_tensors)

  export_path = './BERTmodel/Data/'

但是我收到以下错误：'无法为形状为（（？，）'的张量'input_example_tensor：0'供给形状（）的值

我尝试了以下代码进行预测。

有人可以建议我吗？

pred_sentences = ["The site is great", "I think it's not good"]
def getPrediction(in_sentences):
 labels = ["Negative", "Positive", "Neutral"]
 input_examples = [run_classifier.InputExample(guid="", text_a = x, text_b = None, label = 0)    for x in in_sentences]
 input_features = run_classifier.convert_examples_to_features(input_examples, label_list, MAX_SEQ_LENGTH, tokenizer)
 predict_input_fn = run_classifier.input_fn_builder(features=input_features,   seq_length=MAX_SEQ_LENGTH, is_training=False, drop_remainder=False)
 return predict_input_fn

 from tensorflow.contrib import predictor 
 with tf.Session() as sess:
  predict_fn = predictor.from_saved_model('model_path')
  predictions = predict_fn({"example": getPrediction(pred_sentences)})
  print(predictions)

如何使用tensorflow.contrib重用BERT模型

0 个答案: