gpu support and sklearn support

[BinchaoPeng]

请问现在可以支持gpu了吗?另外是否支持sklearn的gridSearchCV呢？

[IncubatorShokuhou]

@BinchaoPeng 可以自定义estimators和predictor，然后使用支持gpu的评估器作为estimator和predictor。详见https://deep-forest.readthedocs.io/en/stable/advanced_topics/use_customized_estimator.html

[IncubatorShokuhou]

另外，如果你不自定义estimator和predictor的话，cv应该是没问题的。但是如果你用了customized estimators的话，那就会出bug了，因为cv里面会用到sklearn.base.clone (见https://github.com/scikit-learn/scikit-learn/blob/main/sklearn/model_selection/_search.py#L769)

关于clone引起bug的问题，可以看我新提交的issure: #92

[xuyxu]

感谢 @IncubatorShokuhou 的说明。关于使用GPU，可以考虑使用cuml中的随机森林：https://docs.rapids.ai/api/cuml/stable/api.html#random-forest

[BinchaoPeng]

好的，谢谢，我去看看

[xuyxu]

如果准备尝试的话，十分欢迎您在这个issue下面进行反馈 ;-)

[BinchaoPeng]

恩恩，好的，希望这个project可以实现gpu支持和sklearn支持

[BinchaoPeng]

还是报错了，关于GridSearchCV的：

/home/pbc/anaconda3/envs/cuda9/lib/python3.6/site-packages/sklearn/model_selection/_validation.py:700: UserWarning: Scoring failed. The score on this train-test partition for these parameters will be set to nan. Details: 
Traceback (most recent call last):
[CV 1/5] END predictor=xgboost; accuracy: (test=nan) average_precision: (test=nan) f1: (test=nan) roc_auc: (test=nan) total time=10.0min
  File "/home/pbc/anaconda3/envs/cuda9/lib/python3.6/site-packages/sklearn/metrics/_scorer.py", line 56, in _cached_call
    return cache[method]
KeyError: 'decision_function'
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
  File "/home/pbc/anaconda3/envs/cuda9/lib/python3.6/site-packages/sklearn/metrics/_scorer.py", line 334, in _score
    y_pred = method_caller(clf, "decision_function", X)
  File "/home/pbc/anaconda3/envs/cuda9/lib/python3.6/site-packages/sklearn/metrics/_scorer.py", line 58, in _cached_call
    result = getattr(estimator, method)(*args, **kwargs)
AttributeError: 'CascadeForestClassifier' object has no attribute 'decision_function'
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
  File "/home/pbc/anaconda3/envs/cuda9/lib/python3.6/site-packages/sklearn/model_selection/_validation.py", line 687, in _score
    scores = scorer(estimator, X_test, y_test)
  File "/home/pbc/anaconda3/envs/cuda9/lib/python3.6/site-packages/sklearn/metrics/_scorer.py", line 88, in __call__
    *args, **kwargs)
  File "/home/pbc/anaconda3/envs/cuda9/lib/python3.6/site-packages/sklearn/metrics/_scorer.py", line 353, in _score
    y_pred = self._select_proba_binary(y_pred, clf.classes_)
AttributeError: 'CascadeForestClassifier' object has no attribute 'classes_'
  UserWarning,

[xuyxu]

好的，我周末研究一下。

[BinchaoPeng]

class_
decision_funtion
这两个链接或许有用

[BinchaoPeng]

好的，我周末研究一下。

请问你搞定了吗？发现不结合网格搜索和交叉验证这两个功能，不好实用，建议还是接入sklearn比较好

[py666888]

请问不自定义分类器，现在支持网格搜索调参吗？

[BinchaoPeng]

请问不自定义分类器，现在支持网格搜索调参吗？

需要自己实现，不支持sklearn，按照sklearn的方式实现就可以了