- Use `UpdatePosition` for all nodes and skip `FinalizePosition` when external memory is used.
- Create `encode/decode` for node position, this is just as a refactor.
- Reuse code between update position and finalization.
- Support resource view in ellpack.
- Define the CUDA version of MMAP resource.
- Define the CUDA version of malloc resource.
- Refactor cuda runtime API wrappers, and add memory access related wrappers.
- gather windows macros into a single header.
- Use std::uint64_t instead of size_t to avoid implementation-defined type.
- Rename to bst_idx_t, to account for other types of indexing.
- Small cleanup to the base header.
* Cleanup code for distributed training.
- Merge `GetNcclResult` into nccl stub.
- Split up utilities from the main dask module.
- Let Channel return `Result` to accommodate nccl channel.
- Remove old `use_label_encoder` parameter.
This PR adds optional support for loading nccl with `dlopen` as an alternative of compile time linking. This is to address the size bloat issue with the PyPI binary release.
- Add CMake option to load `nccl` at runtime.
- Add an NCCL stub.
After this, `nccl` will be fetched from PyPI when using pip to install XGBoost, either by a user or by `pyproject.toml`. Others who want to link the nccl at compile time can continue to do so without any change.
At the moment, this is Linux only since we only support MNMG on Linux.
- CUDA implementation.
- Extract the broadcasting logic, we will need the context parameter after revamping the collective implementation.
- Some changes to the event loop for fixing a deadlock in CI.
- Move argsort into algorithms.cuh, add support for cuda stream.
* [coll] Pass context to various functions.
In the future, the `Context` object would be required for collective operations, this PR
passes the context object to some required functions to prepare for swapping out the
implementation.
- Rework the precision metric for both CPU and GPU.
- Mention it in the document.
- Cleanup old support code for GPU ranking metric.
- Deterministic GPU implementation.
* Drop support for classification.
* type.
* use batch shape.
* lint.
* cpu build.
* cpu build.
* lint.
* Tests.
* Fix.
* Cleanup error message.
* [CI] Use RAPIDS 22.10
* Store CUDA and RAPIDS versions in one place
* Fix
* Add missing #include
* Update gputreeshap submodule
* Fix
* Remove outdated distributed tests
Support adaptive tree, a feature supported by both sklearn and lightgbm. The tree leaf is recomputed based on residue of labels and predictions after construction.
For l1 error, the optimal value is the median (50 percentile).
This is marked as experimental support for the following reasons:
- The value is not well defined for distributed training, where we might have empty leaves for local workers. Right now I just use the original leaf value for computing the average with other workers, which might cause significant errors.
- Some follow-ups are required, for exact, pruner, and optimization for quantile function. Also, we need to calculate the initial estimation.
* Implement `MaxCategory` in quantile.
* Implement partition-based split for GPU evaluation. Currently, it's based on the existing evaluation function.
* Extract an evaluator from GPU Hist to store the needed states.
* Added some CUDA stream/event utilities.
* Update document with references.
* Fixed a bug in approx evaluator where the number of data points is less than the number of categories.
* Add a new ctor to tensor for `initilizer_list`.
* Change labels from host device vector to tensor.
* Rename the field from `labels_` to `labels` since it's a public member.
* Extend array interface to handle ndarray.
The `ArrayInterface` class is extended to support multi-dim array inputs. Previously this
class handles only 2-dim (vector is also matrix). This PR specifies the expected
dimension at compile-time and the array interface can perform various checks automatically
for input data. Also, adapters like CSR are more rigorous about their input. Lastly, row
vector and column vector are handled without intervention from the caller.
On GPU we use rouding factor to truncate the gradient for deterministic results. This PR changes the gradient representation to fixed point number with exponent aligned with rounding factor.
[breaking] Drop non-deterministic histogram.
Use fixed point for shared memory.
This PR is to improve the performance of GPU Hist.
Co-authored-by: Andy Adinets <aadinets@nvidia.com>
* Use type aliases for discard iterators
* update to include host_vector as thrust 1.12 doesn't bring it in as a side-effect
* cub::DispatchRadixSort requires signed offset types
* Support categorical data for dask functional interface and DQM.
* Implement categorical data support for GPU GK-merge.
* Add support for dask functional interface.
* Add support for DQM.
* Get newer cupy.
