Fix the HierarchicalSigmoidGradOpKernel and refine the codes. Now hsigmoid_op is same with V2 implementation and can pass gradient check.

* change im2sequence for ctc batch inference

* Update im2sequence_op.cc

* change im2sequence for ctc batch inference

* update

* change PR by comment

* fix ocr test error

* fix test_im2sequence

* modify the old name to standard name

* fix test_layers failed

* - Rebase of conv reuse

- clag formatter fixes

- Fix to conv reuse

- Yet another fix

- Fix

- Fix

- clagn format

* - comment update

* Add flag to set prior_box output order.

* Add a fake_quantize_op, which quantize an input tensor to a tensor with lower bits.

* Add region proposal network (RPN) target assign operator and Python API for Faster-RCNN.

* Fix batch norm when only 1 elements in normzalize dimension during training.

1. Fix bug when only true or false branch works.
2. Fix bug in unit testing.

1. Shrink DoubleBufferReader's buffer size to 3.
2. Add BatchReader an option to discard leftover instances.
3. Fix a MultiPassReader bug on pass count.

1. Make PyReader and RandomDataGenerator inherited from FileReader.
2. Remove the memeber variable 'dims_' and realated checks in FileReader.

We can get endpoints of a reader chain.