多机同步时如果reduce_strategy使用fluid.BuildStrategy.ReduceStrategy.Reduce报错
Created by: kolinwei
Traceback (most recent call last):
File "dist_mnist.py", line 163, in <module>
runtime_main(TestDistMnist)
File "/work/weike/svn/baidu/paddle/test/cts_test/dist_base.py", line 243, in runtime_main
model.run_trainer(endpoints, trainer_id, trainers, run_params)
File "/work/weike/svn/baidu/paddle/test/cts_test/dist_base.py", line 176, in run_trainer
build_strategy=build_strategy)
File "/usr/local/lib/python2.7/dist-packages/paddle/fluid/parallel_executor.py", line 165, in __init__
build_strategy, num_trainers, trainer_id)
paddle.fluid.core.EnforceNotMet: can not find the right place for rpc op: recv at [/paddle/paddle/fluid/framework/details/multi_devices_graph_pass.cc:822]
PaddlePaddle Call Stacks: