Created by: guochaorong

1. we use args.use_reader_op, test fail because feeder not defined

2. drop last batch to avoid: File "/opt/python/cp27-cp27mu/lib/python2.7/site-packages/paddle/fluid/parallel_executor.py", line 213, in run feed_tensor_dict) paddle.fluid.core.EnforceNotMet: enforce member_->places_.size() == lod_tensors.size() failed, 8 != 5 The number of samples of current batch is less than the count of devices, currently, it is not