* Fix sparse training for trainer_count=1

For trainer_count=1, the gradient machine is NeuralNetwork, which does not create parameter buf for PARAMETER_GRADIENT for sparse update in Parameter::enableType. But gradient parameter buf is still used in SgdThreadUpdater.

* Minor update to comment

* Fix bug in yield dictionary in DataProvider.
* Also make virtualenv work in Paddle.

* hierarchical rnn document, add new config example

* update inputs_type of label

* add check for unsupported config

* refine hierarchical document

* refine doc title

* update docs, fix paddle to PaddlePaddle

* follow comments

* fix DataProvider create function args bug

Change-Id: I9e3a1c535c805bf30204a14aea8d5143ff534784

* remove PserverForPython.h which is not used

Change-Id: I2b27f1f3c11a42766a92fc689f0f5f1f73ee1d70

* add internal document script

Change-Id: Ia0fec79456caea0b271f9903cc13e8a3d32e0774

* add interface and unittest for nce layer

* follow comments

* support rectangle padding, stride, window and input for PoolProjection

* Follow comments.
1. Remove start
2. refine img_pool_a/b.conf for test_NetworkCompare
3. Split unit test

* Modify the test in img_layers.py

* add interface and unittest of RecurrentGradientMachine for the function of multiple Subsequence inlinks with unequal token length

* some bug fix for sparse matrix

* a minor bug fix

* follow comments to fix bugs

This bug occasionally causes dead lock in test_RecurrentGradientMachine
In general, conditional_variable::notify should be used together with mutex for changing condition.

* reduce trainer count for unit test on MAC OSX

* Correctly handling multiple inputs and integer inputs for recurrent_group

* Fix ScatterAgentLayer for generation

* Revert sequence_(nest)_rnn.conf

* Also refine unittest to multiple iteration to prevent luckily random number.

* remove unnecessary field set in ParameterConfig, Evaluators, etc

* min_pool_size would be infinite by default.
  * add unittest for min_pool_size
* Fix bug in can_over_batch_size
  * add unittest for can_over_batch_size
* Add DEFINE_PROVIDER_EX
* Add default value of should_shuffle
  * When training, the default value of should_shuffle is True.
  * When testing, the default value of should_shuffle is False.
  * User a set a provider should_shuffle or not by pass it to `@provider`
  * should_shuffle can handle a list of value, not just boolean
* Add input order mapping by using name
  * Add unittest
* Add check to check input format.
  * Default is close for speed reason.
  * User could stop train when check error, or continue train without
    this train sample.
* use deque instead of vector in generators pool, make erase
  generator faster.
* Add chinese/english documentation
* Make should shuffle = false in unittest
* Add python files to depends.

Sequences should be sorted according to the number of subsequences they have.

* try to fix bug for ctc_error_evaluator

Change-Id: I8e0a8ea6fc2760652d9c76440a539c90860062d3

* modify RecurrentGradientMachine to support hasSubSeq sequence inlinks with the same number of sentence but different number of tokens for each sentence

Change-Id: Ic71f00a4bb346b4fa93e650dfb4b1a0d8d2338b0

Change-Id: Id66da7b7f5bf9ec80cc19b347e4fb822a5a6f197

* by default, synchronize default_stream after resizeAndCopyFrom
* add sync in some places after resizeAndCopyFrom using other streams

Change cos_sim to use CosSimLayer layer when size=1 and rename convex_comb_layer to linear_comb_layer

Also add an unittest for layers.py, currently for a few things. More need to be added later.

* fix CUDNN_VERSION for backward of CudnnBatchNormLayer

* fix cudnn version number for BatchNorm

* Need PyGuard when set args for calc_batch_size

* Make layer stack trace shows ThreadId, Forward or Backward.

Change-Id: Iba1477adb8c9115c3a67ff2959bb5c878ca706c7