GitLab

* Shut up warnings

- warning: 'void* memset(void*, int, size_t)' clearing an object of non-trivial type 'struct LightGBM::HistogramBinEntry'; use assignment or value-initialization instead [-Wclass-memaccess]
- warning: 'void* memcpy(void*, const void*, size_t)' writing to an object of type 'class std::tuple<int, double, double>' with no trivial copy-assignment; use copy-assignment or copy-initialization instead [-Wclass-memaccess]

* void*

* first draft

* refine a branching

* added code to expose c_api_pred_contrib in the R package

* removed Rprintf

* reverted to previous version of install.libs.R

* minor change

* unix line break

* Read and write datsets from hdfs.
* Only enabled when cmake is run with -DUSE_HDFS:BOOL=TRUE
* Introduces VirtualFile(Reader|Writer) to asbtract VFS differences

* fix the bug

* fix R

* fix the name of custom objective function

* fix multi-class alias

* Update GPU-Windows.rst

* added new aliases for regression_l2 to docs

* fixed new line

* fixed tweedie_variance_power description

* moved gamma and tweedie metrics to other regression metrics

* added new aliases for l2 to docs

* added new alias for l1 to docs

* added new alias for log loss to docs

* added new aliases for multiclass to docs

* added description for default value of metric param

* fixed link to poisson regression and capitalized regressions' names

* removed unused earlier alias type for gamma deviance

* added aliases for multiclass and multiclass_ova

* updated Quick-Start.rst according to Parameters.rst

* added cross-entropy into Features

* added aliases

* unfinished

* some refines.

* fix a bug

* some refines

* remove useless files

* some small fixes

* clean code

* clean code

* fix connect

* fix a broken link

* fix a broken link

* update network.h

Improve training speed in parallel learning where workers is not powers of 2

* update linders_socket.cpp

Improve training speed in parallel learning where workers is not powers of 2

* update linder_topo.cpp

Improve training speed in parallel learning where workers is not powers of 2

* update network.cpp

Improve training speed in parallel learning where workers is not powers of 2

* update linder_topo.cpp

fix a bug

* refine allgather.

* fix a bug.