用于安装 pandas 的 Sagemaker 生命周期配置不起作用答案

【问题标题】：Sagemaker lifecycle configuration for installing pandas not working用于安装 pandas 的 Sagemaker 生命周期配置不起作用
【发布时间】：2020-02-03 10:02:36
【问题描述】：

我正在尝试在生命周期配置中更新熊猫，并且按照 AWS 的示例，我有下一个代码：

#!/bin/bash

set -e

# OVERVIEW
# This script installs a single pip package in a single SageMaker conda environments.

sudo -u ec2-user -i <<EOF
# PARAMETERS
PACKAGE=pandas
ENVIRONMENT=python3
source /home/ec2-user/anaconda3/bin/activate "$ENVIRONMENT"
pip install --upgrade "$PACKAGE"==0.25.3
source /home/ec2-user/anaconda3/bin/deactivate
EOF

然后我把它附加到一个笔记本上，当我进入笔记本并打开一个笔记本文件时，我看到熊猫还没有更新。使用!pip show pandas 我得到：

Name: pandas
Version: 0.24.2
Summary: Powerful data structures for data analysis, time series, and statistics
Home-page: http://pandas.pydata.org
Author: None
Author-email: None
License: BSD
Location: /home/ec2-user/anaconda3/envs/python3/lib/python3.6/site-packages
Requires: pytz, python-dateutil, numpy
Required-by: sparkmagic, seaborn, odo, hdijupyterutils, autovizwidget

所以我们可以看到我确实在 python3 环境中，虽然版本是 0.24。

但是cloudwatch的登录显示已经安装好了：

Collecting pandas==0.25.3 Downloading https://files.pythonhosted.org/packages/52/3f/f6a428599e0d4497e1595030965b5ba455fd8ade6e977e3c819973c4b41d/pandas-0.25.3-cp36-cp36m-manylinux1_x86_64.whl (10.4MB)
2020-02-03T12:33:09.065+01:00
Requirement already satisfied, skipping upgrade: pytz>=2017.2 in ./anaconda3/lib/python3.6/site-packages (from pandas==0.25.3) (2018.4)
2020-02-03T12:33:09.065+01:00
Requirement already satisfied, skipping upgrade: python-dateutil>=2.6.1 in ./anaconda3/lib/python3.6/site-packages (from pandas==0.25.3) (2.7.3)
2020-02-03T12:33:09.065+01:00
Requirement already satisfied, skipping upgrade: numpy>=1.13.3 in ./anaconda3/lib/python3.6/site-packages (from pandas==0.25.3) (1.16.4)
2020-02-03T12:33:09.065+01:00
Requirement already satisfied, skipping upgrade: six>=1.5 in ./anaconda3/lib/python3.6/site-packages (from python-dateutil>=2.6.1->pandas==0.25.3) (1.13.0)
2020-02-03T12:33:09.065+01:00
Installing collected packages: pandas Found existing installation: pandas 0.24.2 Uninstalling pandas-0.24.2: Successfully uninstalled pandas-0.24.2
2020-02-03T12:33:12.066+01:00
Successfully installed pandas-0.25.3

可能是什么问题？

【问题讨论】：

标签： python pandas amazon-web-services pip amazon-sagemaker

【解决方案1】：

如果您只想在 python3 环境中安装软件包，请在您的 Create Sagemaker Lifecycle 配置中使用以下脚本。

#!/bin/bash
sudo -u ec2-user -i <<'EOF'

# This will affect only the Jupyter kernel called "conda_python3".
source activate python3

# Replace myPackage with the name of the package you want to install.
pip install pandas==0.25.3
# You can also perform "conda install" here as well.
source deactivate
EOF

参考：“Lifecycle Configuration Best Practices”

【讨论】：

有没有办法，我们可以在这个脚本中添加requirements.txt？

【解决方案2】：

我遇到了完全相同的问题，即笔记本中没有可用的软件包，而 Lifecycle Cloudwatch 指示特定内核的安装成功。对我有用的解决方案是在打开笔记本之前确保安装完成。

【讨论】：