5 Commits

Author SHA1 Message Date
Rory Mitchell
ef23e424f1 [GPU-Plugin] Add GPU accelerated prediction (#2593)
* [GPU-Plugin] Add GPU accelerated prediction

* Improve allocation message

* Update documentation

* Resolve linker error for predictor

* Add unit tests
2017-08-16 12:31:59 +12:00
PSEUDOTENSOR / Jonathan McKinney
c1104f7d0a [GPU-Plugin] Add throw of asserts and added compute compatibility error check. (#2565)
* [GPU-Plugin] Added compute compatibility error check, added verbose timing
2017-08-10 16:07:07 +12:00
Rory Mitchell
eda9e180f0 [GPU-Plugin] Various fixes (#2579)
* Fix test large

* Add check for max_depth 0

* Update readme

* Add LBS specialisation for dense data

* Add bst_gpair_precise

* Temporarily disable accuracy tests on test_large.py

* Solve unused variable compiler warning

* Fix max_bin > 1024 error
2017-08-05 22:16:23 +12:00
PSEUDOTENSOR / Jonathan McKinney
6b375f6ad8 Multi-threaded XGDMatrixCreateFromMat for faster DMatrix creation (#2530)
* Multi-threaded XGDMatrixCreateFromMat for faster DMatrix creation from numpy arrays for python interface.
2017-07-21 14:43:17 +12:00
PSEUDOTENSOR / Jonathan McKinney
ca7fc9fda3 [GPU-Plugin] Fix gpu_hist to allow matrices with more than just 2^{32} elements. Also fixed CPU hist algorithm. (#2518) 2017-07-18 11:19:27 +12:00