问题
AWS Sagemaker's notebook comes with Scikit-Learn version 0.19.1
I would like to use version 0.20.2. To avoid updating it every time in the notebook code, I tried using the lifecycle configurations. I created one with the following code :
#!/bin/bash
set -e
/home/ec2-user/anaconda3/bin/conda install scikit-learn -y
When I run the attached notebook instance and go to the terminal, the version of scikit-learn found with conda list
is correct (0.20.2). But when I run a notebook and import sklearn, the version is still 0.19.2.
import sklearn
print(sklearn.__version__)
Is there any virtual environment on the SageMaker instances where I should install the package ? How can I fix my notebook lifecycle configuration ?
回答1:
Your conda update does not refer to a specific virtualenv, while your notebook probably does. Therefore you dont see an update on the notebook virtualenv.
来源:https://stackoverflow.com/questions/54184145/aws-sagemaker-does-not-update-the-package