Add flowers dataset for image classification model

Correctly handle print_layer in V2 API

rename CMAKE_CURRENT_LIST_DIR to CMAKE_CURRENT_SOURCE_DIR

update contribute_to_paddle_en.md

images reader: read the data without untarring the tarball file.
image.py: move batch function from reader to image.py

remove duplicated examples, and rename demo to v1_api_demo

Centralize the use of simd intrinsic and implement scalar kernels.

Add doc to the usage of warp-ctc.

Add a Variational Autoencoder (VAE) demo written in V1 API.

print_layer is used to print the values of a layer when evaluating the model for debugging purpose. No layer depends on print_layer. It should be added to the topology if it is used to print some layer already in the topology.

Add GLOBAL option for external library targets

move recordio to github.com/PaddlePaddle/recordio

So that they can be found when paddle is built as a add_subdirectory project in another project.

update docker version

cluster dataset split and reader

Hack rdma cmake build lib

remove unnecessary unit test outputs

Add scale factor for smoothL1 and smoothL1Bp

rename AggregateLevel and ExpandLevel

Fix handling of boot_bias_parameter for recurrent_group in v2 API

Fix size for lstm_step, identity_projection and concat

import 3rd party target in CMake

Delete paddle/dist before build

Modify make_model_diagram.py so that it can be used to draw from proto

Move majel/test/* to majel/; Update generic.cmake