TensorFlow 1.15.0 of Python 3.0 is a built-in component of E-MapReduce (EMR) Data
Science clusters. You can use this component without additional configurations. On
the master node of a Data Science cluster, you can purchase only vCPU resources to
compute TensorFlow jobs. On a core node of a Data Science cluster, you can purchase
vCPU or vGPU resources to compute TensorFlow jobs. This topic describes how to view
the TensorFlow version, switch the TensorFlow version, and install a Python package.
Switch the TensorFlow version
- Download a compressed package that is used to switch the TensorFlow version.
- Use a file transfer tool to upload install_tf_header.tar.gz to a directory of the master node in your Data Science cluster.
Note In this example, the compressed package is uploaded to the /root directory.
- Log on to the master node of your cluster in SSH mode. For more information, see Connect to the master node of an EMR cluster in SSH mode.
- Run the following commands to switch the TensorFlow version:
- Decompress the package.
tar -zxvf install_tf_header.tar.gz
- Switch the TensorFlow version.
- Command syntax
sh install_tf_header.sh <version>
version
specifies the destination version.
- Example: Run the following command to switch the TensorFlow version to 2.0.3:
sh install_tf_header.sh 2.0.3
- Run the pip3 list command to view the TensorFlow version.

The TensorFlow version is switched to 2.0.3.
Install a Python package
- Download a Python package.
- Use a file transfer tool to upload install_app_onds.tar.gz to a directory of the master node in your Data Science cluster.
Note In this example, the package is uploaded to the /root directory.
- Log on to the master node of your cluster in SSH mode. For more information, see Connect to the master node of an EMR cluster in SSH mode.
- Run the following commands to install the Python package on all nodes of your Data
Science cluster:
- Decompress the package.
tar -zxvf install_app_onds.tar.gz
- Install the Python package.