Is input of tf.contrib.cudnn_rnn.CudnnLSTM time-major?

[scotthuang1989]

System information

• Have I written custom code (as opposed to using a stock example script provided in TensorFlow):
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): 17.10
• TensorFlow installed from (source or binary): binary
• TensorFlow version (use command below): 1.6
• Python version: 3.6
• Bazel version (if compiling from source):
• GCC/Compiler version (if compiling from source):
• CUDA/cuDNN version:9/7.0
• GPU model and memory:
• Exact command to reproduce:

Describe the problem

I save this in tensorflow/models: quick_draw exmple

you can see that in the comment of this function. it says that CudnnLSTM is time-major, so it need a transpose. But I don't find this information in documentation here:

Where can I get confirmation of this ?

  def _add_cudnn_rnn_layers(convolved):
    """Adds CUDNN LSTM layers."""
    # Convolutions output [B, L, Ch], while CudnnLSTM is time-major.
    convolved = tf.transpose(convolved, [1, 0, 2])
    lstm = tf.contrib.cudnn_rnn.CudnnLSTM(
        num_layers=params.num_layers,
        num_units=params.num_nodes,
        dropout=params.dropout if mode == tf.estimator.ModeKeys.TRAIN else 0.0,
        direction="bidirectional")
    outputs, _ = lstm(convolved)
    # Convert back from time-major outputs to batch-major outputs.
    outputs = tf.transpose(outputs, [1, 0, 2])
    return outputs

[tensorflowbutler]

Nagging Assignee @shivaniag: It has been 15 days with no activity and this issue has an assignee. Please update the label and/or status accordingly.

[tensorflowbutler]

Nagging Assignee @shivaniag: It has been 16 days with no activity and this issue has an assignee. Please update the label and/or status accordingly.

[xiaodongww]

I also meet this problem and it takes hours for me to figure out that CudnnLSTM is time-major according to this answer.
For a beginner like me, it is also likely to use CudnnLSTM as a substitution of RNNCell and falsely use it as the input of tf.nn.bidirectional_dynamic_rnn.
Would it be possible to clarify the input and output of tf.contrib.cudnn_rnn.CudnnLSTM more clearly in the documentation?
Thanks!