No longer need to use SubsequenceInput. The framework will detect.

User can provide an initializer in ParamAttr. The initializer will be called when at parameters.create().

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

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.

cluster dataset split and reader

Some time we simply want to have a fixed initialization for a model.

* Also, test_plot cannot run in MacOS because of matplotlib & ipython's
  limit.
* Add missing dependency in setup.py.
* fix #2264