Created by: TheodoreG

Hi everyone, I am trying to train model with distributed nccl2 mode. I am following the code of https://github.com/PaddlePaddle/models/tree/develop/fluid/PaddleCV/image_classification/dist_train But when I am trying to create a ParallelExecutor I got following error:

 File "/usr/local/lib/python2.7/dist-packages/paddle/fluid/parallel_executor.py", line 166, in __init__
    build_strategy, num_trainers, trainer_id)
TypeError: __init__(): incompatible constructor arguments. The following argument types are supported:
    1. paddle.fluid.core.ParallelExecutor(arg0: List[paddle.fluid.core.Place], arg1: Set[unicode], arg2: Set[unicode], arg3: paddle.fluid.core.ProgramDesc, arg4: unicode, arg5: paddle.fluid.core.Scope, arg6: List[paddle.fluid.core.Scope], arg7: paddle.fluid.core.ExecutionStrategy, arg8: paddle.fluid.core.BuildStrategy, arg9: int, arg10: int)

Invoked with: [<paddle.fluid.core.Place object at 0x7f1e62872f90>], set([u'fc1_shared_b', u'fc1_shared_w', u'fc2_shared_b', u'fc2_shared_w']), set([u'velocity_2', u'fc2_shared_b', u'fc1_shared_b', u'fc2_shared_w', u'@LR_DECAY_COUNTER@', u'fc1_shared_w', u'velocity_3', u'velocity_0', u'velocity_1']), <paddle.fluid.core.ProgramDesc object at 0x7f1e6283a810>, u'mean_0.tmp_0', <paddle.fluid.core.Scope object at 0x7f1e779e5570>, [], <paddle.fluid.core.ExecutionStrategy object at 0x7f1e62872c00>, <paddle.fluid.core.BuildStrategy object at 0x7f1e62872d20>, '2', '0'

I found that every argument feed into core.ParallelExecutor conformed to the argument type but still got such error.