Created by: gibiansky

I have the following code:

size = 100

boot = paddle.layer.fc(
  size=size,
  act=paddle.activation.Tanh(),
  bias_attr=False,
  input=static_input)

def gru_with_static_input(current_input, static_input):
     mem = paddle.layer.memory(name='memory', size=size, boot_layer=boot)
     gru_inputs = paddle.layer.fc(
         act=paddle.activation.Linear(),
         size=size * 3, bias_attr=False,
         input=[static_input, current_input])
     gru_step = paddle.layer.gru_step(
         name='gru_decoder',
         input=gru_inputs,
         output_mem=mem,
         size=size)
     return gru_step

fwd = paddle.layer.recurrent_group(input=inputs, step=gru_with_static_input, name="fwd")

I get the following error which I don't understand:

  File "paddlepaddle.py", line 117, in lstm
    fwd = paddle.layer.recurrent_group(input=inputs, step=make_gru_step(size, static), name="fwd")
  File "/usr/local/lib/python2.7/dist-packages/paddle/v2/config_base.py", line 52, in wrapped
    out = f(*args, **xargs)
  File "/usr/local/lib/python2.7/dist-packages/paddle/trainer_config_helpers/default_decorators.py", line 53, in __wrapper__
    return func(*args, **kwargs)
  File "/usr/local/lib/python2.7/dist-packages/paddle/trainer_config_helpers/layers.py", line 3643, in recurrent_group
    RecurrentLayerGroupEnd(name=name)
  File "/usr/local/lib/python2.7/dist-packages/paddle/trainer/config_parser.py", line 391, in RecurrentLayerGroupEnd
    layer = g_layer_map[pair.layer_name]
KeyError: u'memory@fwd'

This is using the docker image paddlepaddle/book:latest-gpu.

What is going on? How do I fix this? I took the code from the machine translation notebook and modified it.