cuML RandomForestClassifier: CUDA错误与文档示例

Question

我试图在木星笔记本上运行这个例子，这个例子找到了这里，并复制了下一篇cuML关于分类的介绍--它在6000以下的n_samples中运行良好(这个参数指示生成的数据集的行数)

import cuml
from cuml.datasets.classification import make_classification
from cuml.preprocessing.model_selection import train_test_split
from cuml.ensemble import RandomForestClassifier as cuRF
from sklearn.metrics import accuracy_score
from cupy import asnumpy

# synthetic dataset dimensions
n_samples = 1000
n_features = 10
n_classes = 2

# random forest depth and size
n_estimators = 25
max_depth = 10

# generate synthetic data [ binary classification task ]
X, y = make_classification ( n_classes = n_classes,
                             n_features = n_features,
                             n_samples = n_samples,
                             random_state = 0 )

X_train, X_test, y_train, y_test = train_test_split( X, y, random_state = 0 )

model = cuRF( max_depth = max_depth,
              n_estimators = n_estimators,
              random_state  = 0 )

%time trained_RF = model.fit ( X_train, y_train )

predictions = model.predict ( X_test )

cu_score = cuml.metrics.accuracy_score( y_test, predictions )
sk_score = accuracy_score( asnumpy( y_test ), asnumpy( predictions ) )

在6000以上，我得到了以下CUDA错误和内核崩溃。请注意：

• 将n_features从10增加到5000，而n_samples = 5000运行得非常好。因此，这似乎是数据集的行数，而不是列数的问题。
• 在机器上可用的2个GPU上进行测试(GTX 1050 2GB)
• nvidia-smi显示在运行过程中GPU内存使用率低于25%。
• 库达v11.2
• 驱动程序版本: 460.73.01
• ubuntu 18

任何帮助都是非常感谢的。

数据自动化系统错误：

~/anaconda3/envs/rapids/lib/python3.8/site-packages/cuml/internals/api_decorators.py中的RuntimeError跟踪(最近一次调用)在inner_with_setters(*args，**kwargs) 408 target_val=target_val) 409 -> 410返回函数(*args，**kwargs) 411 412 @ cuml/ensemble/randomforestclassifier.pyx in cuml.ensemble.randomforestclassifier.RandomForestClassifier.fit() RuntimeError： file=/opt/conda/envs/rapids/conda-bld/libcuml_1614210250760/work/cpp/src/decisiontree/quantile/quantile.cuh line=150: call='cub::DeviceRadixSort::SortKeys( (void *)d_temp_RuntimeError->data()，temp_storage_bytes，&d_keys_inbatch_offset，D_key_out->data()，n_sampled_rows，0，8* sizeof(T)，tempmem>stream)‘，在/home/oleg/anaconda3/envs/rapids/lib/python3.8/site-packages/cuml/common/../../../../libcuml++.so(_ZN4raft9exception18collect_call_stackEv+0x46) 0x7fa9b83eef36 #1中，在/home/oleg/anaconda3/envs/rapids/lib/python3.8/site-packages/cuml/common/../../../../libcuml++.so(_ZN4raft10cuda_errorC1ERKNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE+0x69) 0x7fa9b83ef699 #2中获得了64个堆栈帧#0。/home/oleg/anaconda3/envs/rapids/lib/python3.8/site-packages/cuml/common/../../../../libcuml++.so(_ZN2ML12DecisionTree19preprocess_quantileIfiEEvPKT_PKjiiiiSt10shared_ptrI15TemporaryMemoryIS2_T0_EE+0xaaf) 0x7fa9b84Fe7f #3 in /home/oleg/anaconda3/envs/rapids/lib/python3.8/site-packages/cuml/common/../../../../libcuml++.so(_ZN2ML12rfClassifierIfE3fitERKN4raft8handle_tEPKfiiPiiRPNS_20RandomForestMetaDataIfiEE+0xde3) 0x7fa9b8734b63 #4 in /home/oleg/anaconda3 3/envs/急流/lib/python3.8/site-packages/cuml/common/../../../../libcuml++.so(_ZN2ML3fitERKN4raft8handle_tERPNS_20RandomForestMetaDataIfiEEPfiiPiiNS_9RF_paramsEi+0x1fd) 0x7fa9b872f54d #5 in /home/oleg/anaconda3/envs/rapids/lib/python3.8/site-packages/cuml/ensemble/randomforestclassifier.cpython-38-x86_64-linux-gnu.so(+0x3c7e5) 0x7fa98e6d97e5 #6 /home/oleg/anaconda3/envs/rapids/bin/python(PyObject_Call+0x255) 0x5589964052b5 #7 in /home/oleg/anaconda3/envs/rapids/bin/python(_PyEval_EvalFrameDefault+0x21c1) 0x5589964b1de1 #8 /home/oleg/anaconda3/envs/rapids/bin/python(_PyEval_EvalCodeWithName+0x2c3) 0x558996490503 #9 /home/oleg/anaconda3/envs/rapids/bin/python(+0x1b2007) 0x558996492007 #10 /home/oleg/anaconda3/envs/rapids/bin/python(_PyEval_EvalFrameDefault+0x4ca3) 0x5589964b48c3 #11 /home/oleg/anaconda3/envs/rapids/bin/python(_PyEval_EvalCodeWithName+0x2c3) 0x558996490503 #12 /home/oleg/anaconda3 3/envs/rapids/bin/python(PyEval_EvalCodeEx+0x39) 0x558996491559 #13 /home/oleg/anaconda3/envs/rapids/bin/python(PyEval_EvalCode+0x1b) 0x5589965349ab #14 /home/oleg/anaconda3/envs/rapids/bin/python(+0x2731de) 0x5589965531de #15 /home/oleg/anaconda3/envs/rapids/bin/python(+0x128d4b) 0x558996408d4b #16删除/home/oleg/anaconda3/envs/rapids/bin/python(_PyEval_EvalCodeWithName+0x2c3) 0x558996490503 #55，/home/oleg/anaconda3/envs/rapids/bin/python(+0x1b2007) 0x558996492007 #56，/home/oleg/anaconda3/envs/rapids/bin/python(_PyEval_EvalFrameDefault+0x1782) 0x5589964b13a2 #57，/home/oleg/anaconda3/envs/rapids/bin/python(+0x1925da) 0x5589964725da #58，/home/oleg/anaconda3/envs/rapids/bin/python(+0x128d4b) 0x558996408d4b #59，/home/oleg/anaconda3/envs/rapids/bin/python(+0x13b3ea) 0x55899641b3ea #60 /home/oleg/anaconda3/envs/rapids/bin/python(+0x21da4f) 0x5589964fda4f #61 /home/oleg/anaconda3/envs/rapids/bin/python(+0x128fc2) 0x558996408fc2 #62 /home/oleg/anaconda3/envs/rapids/bin/python(_PyEval_EvalFrameDefault+0x92f) 0x5589964b054f #63 /home/oleg/anaconda3/envs/rapids/bin/python(_PyEval_EvalCodeWithName+0x2c3) 0x558996490503

Answer 1

发现这个问题与在cuML中使用RF的实验后端有关，因此在cuRF配置中设置split_algo =0可以通过返回默认后端来解决问题。在编写本报告时，这比使用实验后端慢3倍。