Python User Guide¶

1. Instal¶

We recommend using conda to prepare the environment, especially when running on distributed cluster.

conda create -n zoo python=3.7 # "zoo" is conda enviroment name, you can use any name you like.
conda activate zoo

1.1 Official Release¶

You can install the latest release version of Analytics Zoo as follows:

pip install analytics-zoo

Note: Installing Analytics Zoo will automatically install bigdl==0.12.1, pyspark==2.4.3, conda-pack==0.3.1 and their dependencies.

1.2 Nightly Build¶

<TODO: verify nightly release using pipe –pre>

You can install the latest nightly build of Analytics Zoo as follows:

pip install --pre analytics-zoo

<TODO: maintain a table of nightly build version and add a link here?>

Alternatively, you can find the list of the nightly build versions here, and install a specific version as follows:

pip install analytics-zoo=xxx

2. Run¶

Note: Installing Analytics Zoo from pip will automatically install pyspark. To avoid possible conflicts, you are highly recommended to unset SPARK_HOME if it exists in your environment.

2.1 Interactive Shell¶

You may test if the installation was successful using the interactive Python shell as follows:

2.2 Jupyter Notebook¶

2.3 Python Script¶

3. Compatibility¶

Analytics Zoo has been tested on Python 3.7 with the following library versions: <TODO: add the library list>