Normal prediction with DMatrix is now thread safe with locks. Added inplace prediction is lock free thread safe.
When data is on device (cupy, cudf), the returned data is also on device.
* Implementation for numpy, csr, cudf and cupy.
* Implementation for dask.
* Remove sync in simple dmatrix.
Move this function into gbtree, and uses only updater for doing so. As now the predictor knows exactly how many trees to predict, there's no need for it to update the prediction cache.
* Move prediction cache into Learner.
* Clean-ups
- Remove duplicated cache in Learner and GBM.
- Remove ad-hoc fix of invalid cache.
- Remove `PredictFromCache` in predictors.
- Remove prediction cache for linear altogether, as it's only moving the
prediction into training process but doesn't provide any actual overall speed
gain.
- The cache is now unique to Learner, which means the ownership is no longer
shared by any other components.
* Changes
- Add version to prediction cache.
- Use weak ptr to check expired DMatrix.
- Pass shared pointer instead of raw pointer.
* Simplify DropTrees calling logic
* Add `training` parameter for prediction method.
* [Breaking]: Add `training` to C API.
* Change for R and Python custom objective.
* Correct comment.
Co-authored-by: Philip Hyunsu Cho <chohyu01@cs.washington.edu>
Co-authored-by: Jiaming Yuan <jm.yuan@outlook.com>
* Pass pointer to model parameters.
This PR de-duplicates most of the model parameters except the one in
`tree_model.h`. One difficulty is `base_score` is a model property but can be
changed at runtime by objective function. Hence when performing model IO, we
need to save the one provided by users, instead of the one transformed by
objective. Here we created an immutable version of `LearnerModelParam` that
represents the value of model parameter after configuration.
This PR fixes tree weights in dart being ignored when computing contributions.
* Fix ellpack page source link.
* Add tree weights to compute contribution.
* Extract interaction constraints from split evaluator.
The reason for doing so is mostly for model IO, where num_feature and interaction_constraints are copied in split evaluator. Also interaction constraint by itself is a feature selector, acting like column sampler and it's inefficient to bury it deep in the evaluator chain. Lastly removing one another copied parameter is a win.
* Enable inc for approx tree method.
As now the implementation is spited up from evaluator class, it's also enabled for approx method.
* Removing obsoleted code in colmaker.
They are never documented nor actually used in real world. Also there isn't a single test for those code blocks.
* Unifying the types used for row and column.
As the size of input dataset is marching to billion, incorrect use of int is subject to overflow, also singed integer overflow is undefined behaviour. This PR starts the procedure for unifying used index type to unsigned integers. There's optimization that can utilize this undefined behaviour, but after some testings I don't see the optimization is beneficial to XGBoost.
This makes GPU Hist robust in distributed environment as some workers might not
be associated with any data in either training or evaluation.
* Disable rabit mock test for now: See #5012 .
* Disable dask-cudf test at prediction for now: See #5003
* Launch dask job for all workers despite they might not have any data.
* Check 0 rows in elementwise evaluation metrics.
Using AUC and AUC-PR still throws an error. See #4663 for a robust fix.
* Add tests for edge cases.
* Add `LaunchKernel` wrapper handling zero sized grid.
* Move some parts of allreducer into a cu file.
* Don't validate feature names when the booster is empty.
* Sync number of columns in DMatrix.
As num_feature is required to be the same across all workers in data split
mode.
* Filtering in dask interface now by default syncs all booster that's not
empty, instead of using rank 0.
* Fix Jenkins' GPU tests.
* Install dask-cuda from source in Jenkins' test.
Now all tests are actually running.
* Restore GPU Hist tree synchronization test.
* Check UUID of running devices.
The check is only performed on CUDA version >= 10.x, as 9.x doesn't have UUID field.
* Fix CMake policy and project variables.
Use xgboost_SOURCE_DIR uniformly, add policy for CMake >= 3.13.
* Fix copying data to CPU
* Fix race condition in cpu predictor.
* Fix duplicated DMatrix construction.
* Don't download extra nccl in CI script.
* Use `UpdateAllowUnknown' for non-model related parameter.
Model parameter can not pack an additional boolean value due to binary IO
format. This commit deals only with non-model related parameter configuration.
* Add tidy command line arg for use-dmlc-gtest.
* Move get transpose into cc.
* Clean up headers in host device vector, remove thrust dependency.
* Move span and host device vector into public.
* Install c++ headers.
* Short notes for c and c++.
Co-Authored-By: Philip Hyunsu Cho <chohyu01@cs.washington.edu>
* Refactor configuration [Part II].
* General changes:
** Remove `Init` methods to avoid ambiguity.
** Remove `Configure(std::map<>)` to avoid redundant copying and prepare for
parameter validation. (`std::vector` is returned from `InitAllowUnknown`).
** Add name to tree updaters for easier debugging.
* Learner changes:
** Make `LearnerImpl` the only source of configuration.
All configurations are stored and carried out by `LearnerImpl::Configure()`.
** Remove booster in C API.
Originally kept for "compatibility reason", but did not state why. So here
we just remove it.
** Add a `metric_names_` field in `LearnerImpl`.
** Remove `LazyInit`. Configuration will always be lazy.
** Run `Configure` before every iteration.
* Predictor changes:
** Allocate both cpu and gpu predictor.
** Remove cpu_predictor from gpu_predictor.
`GBTree` is now used to dispatch the predictor.
** Remove some GPU Predictor tests.
* IO
No IO changes. The binary model format stability is tested by comparing
hashing value of save models between two commits
* Only define `gpu_id` and `n_gpus` in `LearnerTrainParam`
* Pass LearnerTrainParam through XGBoost vid factory method.
* Disable all GPU usage when GPU related parameters are not specified (fixes XGBoost choosing GPU over aggressively).
* Test learner train param io.
* Fix gpu pickling.
* Combine thread launches into single launch per tree for gpu_hist
algorithm.
* Address deprecation warning
* Add manual column sampler constructor
* Turn off omp dynamic to get a guaranteed number of threads
* Enable openmp in cuda code
* Upgrade gtest for clang-tidy.
* Use CMake to install GTest instead of mv.
* Don't enforce clang-tidy to return 0 due to errors in thrust.
* Add a small test for tidy itself.
* Reformat.
* Unify logging facilities.
* Enhance `ConsoleLogger` to handle different verbosity.
* Override macros from `dmlc`.
* Don't use specialized gamma when building with GPU.
* Remove verbosity cache in monitor.
* Test monitor.
* Deprecate `silent`.
* Fix doc and messages.
* Fix python test.
* Fix silent tests.
- Improved GPU performance logging
- Only use one execute shards function
- Revert performance regression on multi-GPU
- Use threads to launch NCCL AllReduce
* Multi-GPU support in GPUPredictor.
- GPUPredictor is multi-GPU
- removed DeviceMatrix, as it has been made obsolete by using HostDeviceVector in DMatrix
* Replaced pointers with spans in GPUPredictor.
* Added a multi-GPU predictor test.
* Fix multi-gpu test.
* Fix n_rows < n_gpus.
* Reinitialize shards when GPUSet is changed.
* Tests range of data.
* Remove commented code.
* Remove commented code.
* DMatrix refactor 2
* Remove buffered rowset usage where possible
* Transition to c++11 style iterators for row access
* Transition column iterators to C++ 11
* Replaced std::vector with HostDeviceVector in MetaInfo and SparsePage.
- added distributions to HostDeviceVector
- using HostDeviceVector for labels, weights and base margings in MetaInfo
- using HostDeviceVector for offset and data in SparsePage
- other necessary refactoring
* Added const version of HostDeviceVector API calls.
- const versions added to calls that can trigger data transfers, e.g. DevicePointer()
- updated the code that uses HostDeviceVector
- objective functions now accept const HostDeviceVector<bst_float>& for predictions
* Updated src/linear/updater_gpu_coordinate.cu.
* Added read-only state for HostDeviceVector sync.
- this means no copies are performed if both host and devices access
the HostDeviceVector read-only
* Fixed linter and test errors.
- updated the lz4 plugin
- added ConstDeviceSpan to HostDeviceVector
- using device % dh::NVisibleDevices() for the physical device number,
e.g. in calls to cudaSetDevice()
* Fixed explicit template instantiation errors for HostDeviceVector.
- replaced HostDeviceVector<unsigned int> with HostDeviceVector<int>
* Fixed HostDeviceVector tests that require multiple GPUs.
- added a mock set device handler; when set, it is called instead of cudaSetDevice()
The base margin will need to have length `[num_class] * [number of data points]`.
Otherwise, the array holding prediction results will be only partially
initialized, causing undefined behavior.
Fix: check the length of the base margin. If the length is not correct,
use the global bias (`base_score`) instead. Warn the user about the
substitution.
