Documentation Center

Configuring Helm Chart

Steps to configure Helm Chart on your installation.

The Helm Chart deployment requires that the configuration files match your Kubernetes cluster. For your convenience, we provide sample configuration files that are known to work on Azure and Google GKE respectively.

  • azure.yaml
  • google.yaml

It is recommended that you start with these configuration files and modify them to match your own deployment. At a minimum, you will have to configure which language pairs to use.

Customizing the Values.yaml configuration file

Before you can deploy Language Weaver Edge on your Kubernetes cluster, you need to customize your Values.yaml configuration file. Select one of the provided configuration files based on your current cloud provider or customize your own configuration as per your Kubernetes cluster.

ParameterDefaultComment

google

azure

aws

openshift

falseEnable support for each Kubernetes platform by changing the value to true.
mtedge.image.registryProvide the name of your registry where docker images will be located.
mtedge.ingress.hostDefine the FQDN of your ingress.
mtedge.storage.controller.size10Gi100Gi - 250Gi is recommended for large deployments.
mtedge.storage.controller.classProvide the name of your block storage class.
mtedge.storage.controller.pvcfalseProvide the name of your own PVC (in this case, .controller.size and .controller.class will be ignored).
mtedge.storage.lp.size 100GiAdjust based on the number of language pairs to be installed (count 2-4 GB per language pair).
mtedge.storage.lp.classProvide the name of your filesystem storage class.
mtedge.storage.lp.pvcProvide the name of your own PVC (in this case, .lp.size and .lp.class will be ignored).
mtedge.storage.lp.fsidfalseApplicable only to Amazon EFS when aws: true.
mtedge.license (optional)Upload your license as a file or as a secret.
mtedge.init.admin.secretNameCreate a secret with your Admin username/password.
mtedge.edgeCloudfalseAdd your cloud host, clientID, and secretKey for Edge-Cloud engines.
mtedge.resourcesAdjust min and max resources for controller, translation engine and job engine as per the scale of your deployment.
mtedge.jengines.replicas2Adjust the number of job engines for your deployment. Each job engine will be deployed in a separate pod.
mtedge.tengines0Define the translation engines according to the required language pairs.
mtedge.trainingEngines.replicas0Adjust the number of training engines for your deployment. Each training engine will be deployed in a separate pod and requires a separate GPU on a GPU node.
mtedge.securityContextfalseApplicable only to RedHat OpenShift when openshift: true.

Translation engine configuration

For each language pair to be deployed on the Language Weaver Edge instance, you need to define the translation engine configurations under the tengines: section. It is highly recommended to create placeholder translation engine configurations for each Adapted/Auto Adaptive language pair that you are planning to deploy on the cluster.

Possible translation engine options include:
  • Local language pairs already installed on your language pair storage.
  • Cloud language pairs available on your Language Weaver Cloud subscription.
  • Adaptable language pairs already deployed in your language pair storage or planning to deploy to your language pair storage.
  • Auto Adaptive language pairs that you are planning to deploy to your language pair storage.

Examples

The following are some sample translation engine configurations for the Values.yaml file to be used with different language pair types.

  • Local language pairs
    Syntax
    tengines:
      engspa:
        lp: engspa_generic_srv_tnm_8_6_x_8
        processingUnits: 1
        maxProcessingUnits: 2
        replicas: 1
    
  • Cloud language pairs
    Syntax
    tengines:
      cloud:
        lps:
        - engspa_generic_cloud
        - spaeng_generic_cloud
        - engfra_generic_cloud
        - fraeng_generic_cloud
    
  • Adapted language pairs
    Syntax
    tengines:
      engspa-alp:
        lp: engspa_adapted-<your-ALP-name>_srv_tnm_8_6_x_8
        processingUnits: 1
        maxProcessingUnits: 2
        replicas: 1
    
  • Auto Adaptive language pairs
    Syntax
    tengines:
      engspa-aalp:
        lp: engspa_autoadaptive_srv_tnm
        processingUnits: 1
        maxProcessingUnits: 2
        replicas: 1