```pipdeptree --reverse --packages xgboost```-如何解决库不兼容

如何解决```pipdeptree --reverse --packages xgboost```-如何解决库不兼容

目标

• 我的最终目标是在conda环境中本地运行此kaggle项目。
• 我正在Ubuntu 20.04 LTS，conda环境，python 3.8上运行此程序

输入

• 由于存在库依赖性问题，建议我运行this代码。

pipdeptree --reverse --packages xgboost

输出

Warning!!! Possibly conflicting dependencies found:
* wandb==0.10.4
 - subprocess32 [required: >=3.5.3,installed: ?]
* spyder==4.1.5
 - pyqt5 [required: <5.13,installed: ?]
 - pyqtwebengine [required: <5.13,installed: ?]
* QDarkStyle==2.8.1
 - helpdev [required: >=0.6.10,installed: ?]
* Flask-Compress==1.5.0
 - brotli [required: Any,installed: ?]
* dask-xgboost==0.1.11
 - xgboost [required: <=0.90,installed: 1.2.0]
* dash==1.16.3
 - dash-core-components [required: ==1.12.1,installed: 1.3.1]
 - dash-html-components [required: ==1.1.1,installed: 1.0.1]
 - dash-renderer [required: ==1.8.2,installed: 1.1.2]
 - dash-table [required: ==4.10.1,installed: 4.4.1]
------------------------------------------------------------------------
xgboost==1.2.0
  - dask-xgboost==0.1.11 [requires: xgboost<=0.90]

更新

现在我已经重新安装了XGBoost，并且

conda install -c conda-forge xgboost

当我在jupyter中运行代码导入时，它说

---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
<ipython-input-1-f5a95d2c9926> in <module>
      3 from dask.distributed import Client
      4 from dask import array as da
----> 5 import xgboost as xgb
      6 from xgboost import dask as dxgb
      7 from xgboost.dask import DaskDMatrix

ModuleNotFoundError: No module named 'xgboost'

解决方法

如果创建新的conda环境，是否会遇到相同的错误？例如

conda create -n kaggle_xgboost_test python=3.8 pandas numpy scikit-learn xgboost jupyter
conda activate kaggle_xgboost_test

# Then open a notebook
jupyter-notebook

# And try to import xgboost with e.g.
import xgboost
print("imported ok")

编辑

如果安装/导入的xgboost新环境没有任何问题（如您的注释所暗示），则您遇到的问题是现有环境所特有的。您是否解决了pipdeptree --reverse --packages xgboost所标识的冲突？您尝试过building xgboost from source with GPU support enabled吗？例如

git clone --recursive https://github.com/dmlc/xgboost
cd xgboost
mkdir build
cd build
cmake .. -DUSE_CUDA=ON
make -j4