Ошибка при сохранении модели от tf.compat.v1.train.saver - PullRequest
Новая
       12

Ошибка при сохранении модели от tf.compat.v1.train.saver

0 голосов
/ 19 февраля 2020

Платформа и распространение ОС (например, Linux Ubuntu 16.04): Ubuntu 18.04.3 TensorFlow установлен из: pip Версия TensorFlow: v2.1.0-rc2-17-ge5bf8de Python версия: 3.7.0 от anaconda CUDA / версия cuDNN: 10.1.243

при сохранении модели по 

saver = tf.compat.v1.train.Saver(tf.compat.v1.global_variables(), max_to_keep=3)
path = saver.save(sess, checkpoint_prefix, global_step=current_step)
                            print("Saved model checkpoint to {}".format(path))

возникает ошибка: 

[libprotobuf ERROR google/protobuf/wire_format_lite.cc:577] String field 'tensorflow.TensorShapeProto.Dim.name' contains invalid UTF-8 data when parsing a protocol buffer. Use the 'bytes' type if you intend to send raw bytes.
Traceback (most recent call last):
  File "cnn.py", line 374, in <module>
    train(args, x_train, y_train, vocab_size, x_dev, y_dev)
  File "cnn.py", line 279, in train
    path = saver.save(sess, checkpoint_prefix, global_step=current_step)
  File "xxx/anaconda3/envs/li/lib/python3.7/site-packages/tensorflow_core/python/training/saver.py", line 1203, in save
    save_debug_info=save_debug_info)
  File "xxx/anaconda3/envs/li/lib/python3.7/site-packages/tensorflow_core/python/training/saver.py", line 1246, in export_meta_graph
    graph_def=ops.get_default_graph().as_graph_def(add_shapes=True),
  File "xxx/anaconda3/envs/li/lib/python3.7/site-packages/tensorflow_core/python/framework/ops.py", line 3134, in as_graph_def
    result, _ = self._as_graph_def(from_version, add_shapes)
  File "xxx/anaconda3/envs/li/lib/python3.7/site-packages/tensorflow_core/python/framework/ops.py", line 3051, in _as_graph_def
    graph.ParseFromString(compat.as_bytes(data))
google.protobuf.message.DecodeError: Error parsing message
...