Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Document for device ordinal. (#9398)

Browse Source 
- Rewrite GPU demos. notebook is converted to script to avoid committing additional png plots.
- Add GPU demos into the sphinx gallery.
- Add RMM demos into the sphinx gallery.
- Test for firing threads with different device ordinals.
This commit is contained in:
Jiaming Yuan
2023-07-22 15:26:29 +08:00
committed by GitHub
parent 22b0a55a04
commit 275da176ba
32 changed files with 351 additions and 398 deletions
Download Patch File Download Diff File Expand all files Collapse all files

47
demo/rmm_plugin/README.md
View File

@@ -1,47 +0,0 @@
Using XGBoost with RAPIDS Memory Manager (RMM) plugin (EXPERIMENTAL)
====================================================================
[RAPIDS Memory Manager (RMM)](https://github.com/rapidsai/rmm) library provides a collection of
efficient memory allocators for NVIDIA GPUs. It is now possible to use XGBoost with memory
allocators provided by RMM, by enabling the RMM integration plugin.
The demos in this directory highlights one RMM allocator in particular: **the pool sub-allocator**.
This allocator addresses the slow speed of `cudaMalloc()` by allocating a large chunk of memory
upfront. Subsequent allocations will draw from the pool of already allocated memory and thus avoid
the overhead of calling `cudaMalloc()` directly. See
[this GTC talk slides](https://on-demand.gputechconf.com/gtc/2015/presentation/S5530-Stephen-Jones.pdf)
for more details.
Before running the demos, ensure that XGBoost is compiled with the RMM plugin enabled. To do this,
run CMake with option `-DPLUGIN_RMM=ON` (`-DUSE_CUDA=ON` also required):
```
cmake .. -DUSE_CUDA=ON -DUSE_NCCL=ON -DPLUGIN_RMM=ON
make -j4
```
CMake will attempt to locate the RMM library in your build environment. You may choose to build
RMM from the source, or install it using the Conda package manager. If CMake cannot find RMM, you
should specify the location of RMM with the CMake prefix:
```
# If using Conda:
cmake .. -DUSE_CUDA=ON -DUSE_NCCL=ON -DPLUGIN_RMM=ON -DCMAKE_PREFIX_PATH=$CONDA_PREFIX
# If using RMM installed with a custom location
cmake .. -DUSE_CUDA=ON -DUSE_NCCL=ON -DPLUGIN_RMM=ON -DCMAKE_PREFIX_PATH=/path/to/rmm
```
# Informing XGBoost about RMM pool
When XGBoost is compiled with RMM, most of the large size allocation will go through RMM
allocators, but some small allocations in performance critical areas are using a different
caching allocator so that we can have better control over memory allocation behavior.
Users can override this behavior and force the use of rmm for all allocations by setting
the global configuration ``use_rmm``:
``` python
with xgb.config_context(use_rmm=True):
clf = xgb.XGBClassifier(tree_method="gpu_hist")
```
Depending on the choice of memory pool size or type of allocator, this may have negative
performance impact.
* [Using RMM with a single GPU](./rmm_singlegpu.py)
* [Using RMM with a local Dask cluster consisting of multiple GPUs](./rmm_mgpu_with_dask.py)

51
demo/rmm_plugin/README.rst Normal file
View File

@@ -0,0 +1,51 @@
Using XGBoost with RAPIDS Memory Manager (RMM) plugin (EXPERIMENTAL)
====================================================================
`RAPIDS Memory Manager (RMM) <https://github.com/rapidsai/rmm>`__ library provides a
collection of efficient memory allocators for NVIDIA GPUs. It is now possible to use
XGBoost with memory allocators provided by RMM, by enabling the RMM integration plugin.
The demos in this directory highlights one RMM allocator in particular: **the pool
sub-allocator**. This allocator addresses the slow speed of ``cudaMalloc()`` by
allocating a large chunk of memory upfront. Subsequent allocations will draw from the pool
of already allocated memory and thus avoid the overhead of calling ``cudaMalloc()``
directly. See `this GTC talk slides
<https://on-demand.gputechconf.com/gtc/2015/presentation/S5530-Stephen-Jones.pdf>`_ for
more details.
Before running the demos, ensure that XGBoost is compiled with the RMM plugin enabled. To do this,
run CMake with option ``-DPLUGIN_RMM=ON`` (``-DUSE_CUDA=ON`` also required):
.. code-block:: sh
cmake .. -DUSE_CUDA=ON -DUSE_NCCL=ON -DPLUGIN_RMM=ON
make -j$(nproc)
CMake will attempt to locate the RMM library in your build environment. You may choose to build
RMM from the source, or install it using the Conda package manager. If CMake cannot find RMM, you
should specify the location of RMM with the CMake prefix:
.. code-block:: sh
# If using Conda:
cmake .. -DUSE_CUDA=ON -DUSE_NCCL=ON -DPLUGIN_RMM=ON -DCMAKE_PREFIX_PATH=$CONDA_PREFIX
# If using RMM installed with a custom location
cmake .. -DUSE_CUDA=ON -DUSE_NCCL=ON -DPLUGIN_RMM=ON -DCMAKE_PREFIX_PATH=/path/to/rmm
********************************
Informing XGBoost about RMM pool
********************************
When XGBoost is compiled with RMM, most of the large size allocation will go through RMM
allocators, but some small allocations in performance critical areas are using a different
caching allocator so that we can have better control over memory allocation behavior.
Users can override this behavior and force the use of rmm for all allocations by setting
the global configuration ``use_rmm``:
.. code-block:: python
with xgb.config_context(use_rmm=True):
clf = xgb.XGBClassifier(tree_method="hist", device="cuda")
Depending on the choice of memory pool size or type of allocator, this may have negative
performance impact.

40
demo/rmm_plugin/rmm_mgpu_with_dask.py
View File

@@ -1,3 +1,7 @@
"""
Using rmm with Dask
===================
"""
import dask
from dask.distributed import Client
from dask_cuda import LocalCUDACluster
@@ -11,25 +15,33 @@ def main(client):
# xgb.set_config(use_rmm=True)
X, y = make_classification(n_samples=10000, n_informative=5, n_classes=3)
# In pratice one should prefer loading the data with dask collections instead of using
# `from_array`.
# In pratice one should prefer loading the data with dask collections instead of
# using `from_array`.
X = dask.array.from_array(X)
y = dask.array.from_array(y)
dtrain = xgb.dask.DaskDMatrix(client, X, label=y)
params = {'max_depth': 8, 'eta': 0.01, 'objective': 'multi:softprob', 'num_class': 3,
'tree_method': 'gpu_hist', 'eval_metric': 'merror'}
output = xgb.dask.train(client, params, dtrain, num_boost_round=100,
evals=[(dtrain, 'train')])
bst = output['booster']
history = output['history']
for i, e in enumerate(history['train']['merror']):
print(f'[{i}] train-merror: {e}')
params = {
"max_depth": 8,
"eta": 0.01,
"objective": "multi:softprob",
"num_class": 3,
"tree_method": "hist",
"eval_metric": "merror",
"device": "cuda",
}
output = xgb.dask.train(
client, params, dtrain, num_boost_round=100, evals=[(dtrain, "train")]
)
bst = output["booster"]
history = output["history"]
for i, e in enumerate(history["train"]["merror"]):
print(f"[{i}] train-merror: {e}")
if __name__ == '__main__':
# To use RMM pool allocator with a GPU Dask cluster, just add rmm_pool_size option to
# LocalCUDACluster constructor.
with LocalCUDACluster(rmm_pool_size='2GB') as cluster:
if __name__ == "__main__":
# To use RMM pool allocator with a GPU Dask cluster, just add rmm_pool_size option
# to LocalCUDACluster constructor.
with LocalCUDACluster(rmm_pool_size="2GB") as cluster:
with Client(cluster) as client:
main(client)

7
demo/rmm_plugin/rmm_singlegpu.py
View File

@@ -1,3 +1,7 @@
"""
Using rmm on a single node device
=================================
"""
import rmm
from sklearn.datasets import make_classification
@@ -16,7 +20,8 @@ params = {
"eta": 0.01,
"objective": "multi:softprob",
"num_class": 3,
"tree_method": "gpu_hist",
"tree_method": "hist",
"device": "cuda",
}
# XGBoost will automatically use the RMM pool allocator
bst = xgb.train(params, dtrain, num_boost_round=100, evals=[(dtrain, "train")])