AutoGluonをGoogle Colab GPU環境で実行する方法

AutoGluonとは

mxnet製のAuto MLライブラリ.
https://autogluon.mxnet.io/

Setup on Linux

公式ページより

# Here we assume CUDA 10.0 is installed.  You should change the number
# according to your own CUDA version (e.g. mxnet-cu101 for CUDA 10.1).
pip install --upgrade mxnet-cu100
pip install autogluon

Google Colabで実行

pip install --upgrade mxnet-cu100
pip install autogluon
from autogluon import TabularPrediction as task

ImportError Traceback (most recent call last)
in ()
1 get_ipython().system('pip install mxnet autogluon')
----> 2 from autogluon import TabularPrediction as task

怒られる.
CPU利用でも同様.

解決

色々試したところ、下記の方法で解決.
https://github.com/awslabs/autogluon/issues/163

!pip uninstall -y mkl
!pip install --upgrade mxnet-cu100
!pip install autogluon
!pip install -U ipykernel
# and restart runtime

今後colabのGPUやcudaが変わるとまたエラーするかもしれない.