[TVM Tensorflow frontend] Input information could not be found in the node

Troubleshooting
#1

When compiling tensorflow dssm model, I get the following error：

Traceback (most recent call last):

  File "from_tensorflow.py", line 114, in <module>
    shape=shape_dict)

  File "/usr1/zmh/tvm/python/tvm/relay/frontend/tensorflow.py", line 2715, in from_tensorflow
    mod, params = g.from_tensorflow(graph, layout, shape, outputs)

  File "/usr1/zmh/tvm/python/tvm/relay/frontend/tensorflow.py", line 2333, in from_tensorflow
    control_flow_node_map)

  File "/usr1/zmh/tvm/python/tvm/relay/frontend/tensorflow.py", line 2620, in _convert_control_flow_operator
    op = self._nodes[node.input[0]]

KeyError: 'RNN/word/rnn/cond/word/LSTMBlockCell_12:6'

The node information where the conversion error occurs is as follows

, name: "pred_rnn/rnn/cond/cond/Switch_1"
op: "Switch"
input: "RNN/word/rnn/cond/word/LSTMBlockCell_12:6"
input: "pred_rnn/rnn/cond/cond/pred_id"
attr {
  key: "T"
  value {
    type: DT_FLOAT
  }
}
attr {
  key: "_class"
  value {
    list {
      s: "loc:@RNN/word/rnn/cond/word/LSTMBlockCell_12"
    }
  }
}

env info

tf version: 1.12

It seems that the input node cannot be located when the operator is converted

Whether the output result index is not supported

Thanks in advance for the help of experts