Created by: CrossLee1

I make a test for the layer recurrent_group following this. My net config is as follwings:

input = data_layer(name='feat', size=feat_size) label = data_layer(name='label', size=label_size) def step_test(input): output = fc_layer(input=input, size=1024, act=LinearActivation(), bias_attr=False) return output decoder = recurrent_group(name='group', step=step_test, input=[StaticInput(input=input, is_seq=True)]) last = last_seq(decoder) output = fc_layer(input=last, size=label_size, name='output', bias_attr=bias_attr, act=SigmoidActivation())

and my slots is obj.slots = [DenseSlot(obj.ftr_dim), SparseNonValueSlot(obj.label_size)]

When I begin to training, error occurs: Current Layer forward/backward stack is LayerName: group LayerName: label LayerName: feat *** Aborted at 1478511490 (unix time) try "date -d @1478511490" if you are using GNU date *** Current Layer forward/backward stack is PC: @ 0x93a62d paddle::RecurrentGradientMachine::forward() Current Layer forward/backward stack is *** SIGSEGV (@0x8) received by PID 29766 (TID 0x7f44921fd700) from PID 8; stack trace: *** Current Layer forward/backward stack is @ 0x7f44b74c3160 (unknown) Current Layer forward/backward stack is @ 0x93a62d paddle::RecurrentGradientMachine::forward() Current Layer forward/backward stack is @ 0x92fde3 paddle::RecurrentLayerGroup::forward() Current Layer forward/backward stack is @ 0x9103ac paddle::NeuralNetwork::forward() Current Layer forward/backward stack is @ 0x90162f paddle::TrainerThread::forward() Current Layer forward/backward stack is @ 0x902835 paddle::TrainerThread::computeThread() Current Layer forward/backward stack is @ 0x7f44b6d5a8a0 execute_native_thread_routine Current Layer forward/backward stack is @ 0x7f44b74bb1c3 start_thread Current Layer forward/backward stack is @ 0x7f44b67ce12d __clone

What is the problem? Thank~~