Configure runtime environment for document classification
To enable the document classification capability for the catalog source, you must configure an elastic runtime environment and configure at least one GPU node on the runtime environment.
A GPU node is a prerequisite to train the classification engine and run the document classification capability. For better performance during classification training and inference, you can install more than one GPU node.
1Configure an elastic runtime environment in Administrator.
For information about how to configure and use the elastic runtime environment, see Elastic Runtime Environments in the Administrator help.
2Download the gpu_setup.zip file.
The ZIP file is available in the following location:
3Extract the ZIP file on the master node of your elastic runtime environment installation.
Run the following command:
unzip gpu_setup.zip -d gpu_setup && cd gpu_setup
4From the extracted folder, open the gpu_node.txt file and configure the following key-value pairs:
Key
Value
JOB_NODE_ASG_NAME
Name of the existing job-node Auto Scaling Group for your elastic runtime environment cluster.
SSH_KEY_PATH
The absolute path to the PEM key that was generated during the elastic runtime cluster installation.
SSH_USER
The SSH user name for connecting to the cluster nodes. Default is cloud-user.
GPU_WORKER_INSTANCE_TYPE
The EC2 instance type for GPU nodes. Default is g5.xlarge. The minimum requirement is g4dn.xlarge.
5Run the preflight check.
Use the following command:
./gpu-setup-preflight.sh
Ensure that there are no failures in the summary. Review the warnings.
6Run the installer to install the GPU node on the runtime environment.
Use the following command:
./gpu-setup-master-multi.sh
Optionally, add the --count=N parameter where N indicates the desired number of GPU nodes to provision.
After you configure an elastic runtime environment with GPU nodes, select this runtime environment to run the classification engine training jobs in Metadata Command Center. For more information, see Enable document classification.
Enable document classification
After configuring an elastic runtime environment in Administrator, you must enable document classification and select the runtime environment to run the document classification training jobs in Metadata Command Center.
To enable document classification in your organization, perform the following steps:
1In Metadata Command Center, go to the Configure page.
2Click the Document Classification tab.
3Select Enable Document Classification.
4Select the elastic runtime environment that you configured with GPU nodes in Administrator.
5Click Save.
You can now enable the document classification capability for the catalog source to automatically identify and classify unstructured documents, such as PDFs and text files.