Bitnami package for DeepSpeed

DeepSpeed is deep learning software suite for empowering ChatGPT-like model training. Features dense or sparse model inference, high throughput and high compression.

Overview of DeepSpeed

Trademarks: This software listing is packaged by Bitnami. The respective trademarks mentioned in the offering are owned by the respective companies, and use of them does not imply any affiliation or endorsement.

TL;DR

helm install my-release oci://REGISTRY_NAME/REPOSITORY_NAME/deepspeed

Note: You need to substitute the placeholders REGISTRY_NAME and REPOSITORY_NAME with a reference to your Helm chart registry and repository.

Introduction

This chart bootstraps a DeepSpeed deployment on a Kubernetes cluster using the Helm package manager.

Python is built for full integration into Python that enables you to use it with its libraries and main packages.

Bitnami charts can be used with Kubeapps for deployment and management of Helm Charts in clusters.

Prerequisites

• Kubernetes 1.23+
• Helm 3.8.0+
• PV provisioner support in the underlying infrastructure

Installing the Chart

To install the chart with the release name my-release:

helm install my-release oci://REGISTRY_NAME/REPOSITORY_NAME/deepspeed

Note: You need to substitute the placeholders REGISTRY_NAME and REPOSITORY_NAME with a reference to your Helm chart registry and repository. For example, in the case of Bitnami, you need to use REGISTRY_NAME=registry-1.docker.io and REPOSITORY_NAME=bitnamicharts.

These commands deploy DeepSpeed on the Kubernetes cluster in the default configuration. The Parameters section lists the parameters that can be configured.

Tip: List all releases using helm list

Uninstalling the Chart

To uninstall/delete the my-release deployment:

helm delete my-release

The command removes all the Kubernetes components associated with the chart and deletes the release.

Configuration and installation details

Resource requests and limits

Bitnami charts allow setting resource requests and limits for all containers inside the chart deployment. These are inside the resources value (check parameter table). Setting requests is essential for production workloads and these should be adapted to your specific use case.

To make this process easier, the chart contains the resourcesPreset values, which automatically sets the resources section according to different presets. Check these presets in the bitnami/common chart. However, in production workloads using resourcePreset is discouraged as it may not fully adapt to your specific needs. Find more information on container resource management in the official Kubernetes documentation.

Rolling VS Immutable tags

It is strongly recommended to use immutable tags in a production environment. This ensures your deployment does not change automatically if the same tag is updated with a different image.

Bitnami will release a new chart updating its containers if a new version of the main container, significant changes, or critical vulnerabilities exist.

Deploy as Job

By default, the chart will deploy the client container (the one that connects to the Deepspeed workers) as a Deployment. This allows you to enter the container via kubectl exec and perform operations. In case you want to deploy it as a Kubernetes job, set the client.useJob=true value.

Loading your files

The DeepSpeed chart supports three different ways to load your files. In order of priority, they are:

1. Existing config map
2. Add files in the values.yaml
3. Cloning a git repository

This means that if you specify a config map with your files, it won’t check the files defined in values.yaml directory nor the git repository.

In order to use an existing config map, set the source.existingConfigMap=my-config-map parameter.

To add your files in the values.yaml file, set the source.configmap object with the files.

Finally, if you want to clone a git repository you can use those parameters:

source.type=git
source.git.repository=https://github.com/my-user/oci://REGISTRY_NAME/REPOSITORY_NAME
source.git.revision=master

Setting Pod’s affinity

This chart allows you to set your custom affinity using the affinity parameter. Find more information about Pod’s affinity in the kubernetes documentation.

As an alternative, you can use of the preset configurations for pod affinity, pod anti-affinity, and node affinity available at the bitnami/common chart. To do so, set the podAffinityPreset, podAntiAffinityPreset, or nodeAffinityPreset parameters.

Additional environment variables

In case you want to add extra environment variables (useful for advanced operations like custom init scripts), you can use the extraEnvVars property inside each of the subsections: client, worker.

client:
  extraEnvVars:
    - name: LOG_LEVEL
      value: error

worker:
  extraEnvVars:
    - name: LOG_LEVEL
      value: error

Alternatively, you can use a ConfigMap or a Secret with the environment variables. To do so, use the extraEnvVarsCM or the extraEnvVarsSecret values.

Sidecars

If additional containers are needed in the same pod as Milvus (such as additional metrics or logging exporters), they can be defined using the sidecars parameter.

sidecars:
- name: your-image-name
  image: your-image
  imagePullPolicy: Always
  ports:
  - name: portname
    containerPort: 1234

If these sidecars export extra ports, extra port definitions can be added using the service.extraPorts parameter (where available), as shown in the example below:

service:
  extraPorts:
  - name: extraPort
    port: 11311
    targetPort: 11311

NOTE: This Helm chart already includes sidecar containers for the Prometheus exporters (where applicable). These can be activated by adding the --enable-metrics=true parameter at deployment time. The sidecars parameter should therefore only be used for any extra sidecar containers.

If additional init containers are needed in the same pod, they can be defined using the initContainers parameter. Here is an example:

initContainers:
  - name: your-image-name
    image: your-image
    imagePullPolicy: Always
    ports:
      - name: portname
        containerPort: 1234

Learn more about sidecar containers and init containers.

Persistence

The Bitnami DeepSpeed image can persist data. If enabled, the persisted path is /bitnami/deepspeed/data by default.

The chart mounts a Persistent Volume at this location. The volume is created using dynamic volume provisioning.

Adjust permissions of persistent volume mountpoint

As the image run as non-root by default, it is necessary to adjust the ownership of the persistent volume so that the container can write data into it.

By default, the chart is configured to use Kubernetes Security Context to automatically change the ownership of the volume. However, this feature does not work in all Kubernetes distributions. As an alternative, this chart supports using an initContainer to change the ownership of the volume before mounting it in the final destination.

You can enable this initContainer by setting volumePermissions.enabled to true.

Parameters

Global parameters

Name	Description	Value
global.imageRegistry	Global Docker image registry	""
global.imagePullSecrets	Global Docker registry secret names as an array	[]
global.storageClass	Global StorageClass for Persistent Volume(s)	""
global.compatibility.openshift.adaptSecurityContext	Adapt the securityContext sections of the deployment to make them compatible with Openshift restricted-v2 SCC: remove runAsUser, runAsGroup and fsGroup and let the platform use their allowed default IDs. Possible values: auto (apply if the detected running cluster is Openshift), force (perform the adaptation always), disabled (do not perform adaptation)	auto

Common parameters

Name	Description	Value
kubeVersion	Override Kubernetes version	""
nameOverride	String to partially override common.names.fullname	""
fullnameOverride	String to fully override common.names.fullname	""
commonLabels	Labels to add to all deployed objects	{}
commonAnnotations	Annotations to add to all deployed objects	{}
clusterDomain	Kubernetes cluster domain name	cluster.local
extraDeploy	Array of extra objects to deploy with the release	[]
diagnosticMode.enabled	Enable diagnostic mode (all probes will be disabled and the command will be overridden)	false
diagnosticMode.command	Command to override all containers in the deployments/statefulsets	["sleep"]
diagnosticMode.args	Args to override all containers in the deployments/statefulsets	["infinity"]

Source code parameters

Name	Description	Value
image.registry	Deepspeed image registry	REGISTRY_NAME
image.repository	Deepspeed image repository	REPOSITORY_NAME/deepspeed
image.digest	Deepspeed image digest in the way sha256:aa…. Please note this parameter, if set, will override the tag	""
image.pullPolicy	Deepspeed image pull policy	IfNotPresent
image.pullSecrets	Specify docker-registry secret names as an array	[]
source.type	Where the source comes from: Possible values: configmap, git, custom	configmap
source.launchCommand	deepspeed command to run over the project	""
source.configMap	List of files of the project	{}
source.existingConfigMap	Name of a configmap containing the files of the project	""
source.git.repository	Repository that holds the files	""
source.git.revision	Revision from the repository to checkout	""
source.git.extraVolumeMounts	Add extra volume mounts for the Git container	[]
config.defaultHostFile	Host file generated by default (only edit if you know what you are doing)	""
config.overrideHostFile	Override default host file with the content in this value	""
config.existingHostFileConfigMap	Name of a ConfigMap containing the hostfile	""
config.defaultSSHClient	Default SSH client configuration for the client node (only edit if you know what you are doing)	""
config.overrideSSHClient	Override default SSH cliient configuration with the content in this value	""
config.existingSSHClientConfigMap	Name of a ConfigMap containing the SSH client configuration	""
config.defaultSSHServer	Default SSH Server configuration for the worker nodes (only edit if you know what you are doing)	""
config.overrideSSHServer	Overidde SSH Server configuration with the content in this value	""
config.existingSSHServerConfigMap	Name of a ConfigMap with with the SSH Server configuration	""
config.sshPrivateKey	Private key for the client node to connect to the worker nodes	""
config.existingSSHKeySecret	Name of a secret containing the ssh private key	""

Client Deployment Parameters

Name	Description	Value
client.enabled	Enable Client deployment	true
client.useJob	Deploy as job	false
client.backoffLimit	set backoff limit of the job	10
client.extraEnvVars	Array with extra environment variables to add to client nodes	[]
client.extraEnvVarsCM	Name of existing ConfigMap containing extra env vars for client nodes	""
client.extraEnvVarsSecret	Name of existing Secret containing extra env vars for client nodes	""
client.annotations	Annotations for the client deployment	{}
client.command	Override default container command (useful when using custom images)	[]
client.args	Override default container args (useful when using custom images)	[]
client.terminationGracePeriodSeconds	Client termination grace period (in seconds)	""
client.livenessProbe.enabled	Enable livenessProbe on Client nodes	true
client.livenessProbe.initialDelaySeconds	Initial delay seconds for livenessProbe	5
client.livenessProbe.periodSeconds	Period seconds for livenessProbe	30
client.livenessProbe.timeoutSeconds	Timeout seconds for livenessProbe	20
client.livenessProbe.failureThreshold	Failure threshold for livenessProbe	5
client.livenessProbe.successThreshold	Success threshold for livenessProbe	1
client.readinessProbe.enabled	Enable readinessProbe on Client nodes	true
client.readinessProbe.initialDelaySeconds	Initial delay seconds for readinessProbe	5
client.readinessProbe.periodSeconds	Period seconds for readinessProbe	30
client.readinessProbe.timeoutSeconds	Timeout seconds for readinessProbe	30
client.readinessProbe.failureThreshold	Failure threshold for readinessProbe	5
client.readinessProbe.successThreshold	Success threshold for readinessProbe	1
client.startupProbe.enabled	Enable startupProbe on Client containers	false
client.startupProbe.initialDelaySeconds	Initial delay seconds for startupProbe	5
client.startupProbe.periodSeconds	Period seconds for startupProbe	30
client.startupProbe.timeoutSeconds	Timeout seconds for startupProbe	5
client.startupProbe.failureThreshold	Failure threshold for startupProbe	5
client.startupProbe.successThreshold	Success threshold for startupProbe	1
client.customLivenessProbe	Custom livenessProbe that overrides the default one	{}
client.customReadinessProbe	Custom readinessProbe that overrides the default one	{}
client.customStartupProbe	Custom startupProbe that overrides the default one	{}
client.resourcesPreset	Set container resources according to one common preset (allowed values: none, nano, micro, small, medium, large, xlarge, 2xlarge). This is ignored if client.resources is set (client.resources is recommended for production).	medium
client.resources	Set container requests and limits for different resources like CPU or memory (essential for production workloads)	{}
client.podSecurityContext.enabled	Enabled Client pods’ Security Context	true
client.podSecurityContext.fsGroupChangePolicy	Set filesystem group change policy	Always
client.podSecurityContext.sysctls	Set kernel settings using the sysctl interface	[]
client.podSecurityContext.supplementalGroups	Set filesystem extra groups	[]
client.podSecurityContext.fsGroup	Set Client pod’s Security Context fsGroup	1001
client.containerSecurityContext.enabled	Enabled Client containers’ Security Context	true
client.containerSecurityContext.seLinuxOptions	Set SELinux options in container	{}
client.containerSecurityContext.runAsUser	Set Client containers’ Security Context runAsUser	1001
client.containerSecurityContext.runAsGroup	Set Client containers’ Security Context runAsGroup	1001
client.containerSecurityContext.runAsNonRoot	Set Client containers’ Security Context runAsNonRoot	true
client.containerSecurityContext.readOnlyRootFilesystem	Set Client containers’ Security Context runAsNonRoot	true
client.containerSecurityContext.privileged	Set Client containers’ Security Context privileged	false
client.containerSecurityContext.allowPrivilegeEscalation	Set Client container’s privilege escalation	false
client.containerSecurityContext.capabilities.drop	Set Client container’s Security Context runAsNonRoot	["ALL"]
client.containerSecurityContext.seccompProfile.type	Set Client container’s Security Context seccomp profile	RuntimeDefault
client.lifecycleHooks	for the client container(s) to automate configuration before or after startup	{}
client.runtimeClassName	Name of the runtime class to be used by pod(s)	""
client.automountServiceAccountToken	Mount Service Account token in pod	false
client.hostAliases	client pods host aliases	[]
client.labels	Extra labels for the client deployment	{}
client.podLabels	Extra labels for client pods	{}
client.podAnnotations	Annotations for client pods	{}
client.podAffinityPreset	Pod affinity preset. Ignored if client.affinity is set. Allowed values: soft or hard	""
client.podAntiAffinityPreset	Pod anti-affinity preset. Ignored if client.affinity is set. Allowed values: soft or hard	soft
client.nodeAffinityPreset.type	Node affinity preset type. Ignored if client.affinity is set. Allowed values: soft or hard	""
client.nodeAffinityPreset.key	Node label key to match. Ignored if client.affinity is set	""
client.nodeAffinityPreset.values	Node label values to match. Ignored if client.affinity is set	[]
client.affinity	Affinity for Client pods assignment	{}
client.nodeSelector	Node labels for Client pods assignment	{}
client.tolerations	Tolerations for Client pods assignment	[]
client.topologySpreadConstraints	Topology Spread Constraints for pod assignment spread across your cluster among failure-domains	[]
client.priorityClassName	Client pods’ priorityClassName	""
client.schedulerName	Kubernetes pod scheduler registry	""
client.updateStrategy.type	Client statefulset strategy type	RollingUpdate
client.updateStrategy.rollingUpdate	Client statefulset rolling update configuration parameters	{}
client.extraVolumes	Optionally specify extra list of additional volumes for the Client pod(s)	[]
client.extraVolumeMounts	Optionally specify extra list of additional volumeMounts for the Client container(s)	[]
client.sidecars	Add additional sidecar containers to the Client pod(s)	[]
client.enableDefaultInitContainers	Deploy default init containers	true
client.initContainers	Add additional init containers to the Client pod(s)	[]
client.networkPolicy.enabled	Enable creation of NetworkPolicy resources	true
client.networkPolicy.allowExternalEgress	Allow the pod to access any range of port and all destinations.	true
client.networkPolicy.extraIngress	Add extra ingress rules to the NetworkPolicy	[]
client.networkPolicy.extraEgress	Add extra ingress rules to the NetworkPolicy	[]
client.serviceAccount.create	Enable creation of ServiceAccount for Client pods	true
client.serviceAccount.name	The name of the ServiceAccount to use	""
client.serviceAccount.automountServiceAccountToken	Allows auto mount of ServiceAccountToken on the serviceAccount created	false
client.serviceAccount.annotations	Additional custom annotations for the ServiceAccount	{}

Deepspeed Client persistence paramaters

Name	Description	Value
client.persistence.enabled	Use a PVC to persist data	false
client.persistence.storageClass	discourse & sidekiq data Persistent Volume Storage Class	""
client.persistence.existingClaim	Use a existing PVC which must be created manually before bound	""
client.persistence.mountPath	Path to mount the volume at	/bitnami/deepspeed/data
client.persistence.accessModes	Persistent Volume Access Mode	["ReadWriteOnce"]
client.persistence.dataSource	Custom PVC data source	{}
client.persistence.selector	Selector to match an existing Persistent Volume for the client data PVC	{}
client.persistence.size	Size of data volume	8Gi
client.persistence.labels	Persistent Volume labels	{}
client.persistence.annotations	Persistent Volume annotations	{}
client.pdb.create	Enable/disable a Pod Disruption Budget creation	true
client.pdb.minAvailable	Minimum number/percentage of pods that should remain scheduled	""
client.pdb.maxUnavailable	Maximum number/percentage of pods that may be made unavailable. Defaults to 1 if both client.pdb.minAvailable and client.pdb.maxUnavailable are empty.	""

Worker Deployment Parameters

Name	Description	Value
worker.enabled	Enable Worker deployment	true
worker.slotsPerNode	Number of slots available per worker node	1
worker.extraEnvVars	Array with extra environment variables to add to client nodes	[]
worker.extraEnvVarsCM	Name of existing ConfigMap containing extra env vars for client nodes	""
worker.extraEnvVarsSecret	Name of existing Secret containing extra env vars for client nodes	""
worker.command	Override default container command (useful when using custom images)	[]
worker.args	Override default container args (useful when using custom images)	[]
worker.replicaCount	Number of Worker replicas to deploy	3
worker.terminationGracePeriodSeconds	Worker termination grace period (in seconds)	""
worker.containerPorts.ssh	SSH port for Worker	2222
worker.livenessProbe.enabled	Enable livenessProbe on Worker nodes	true
worker.livenessProbe.initialDelaySeconds	Initial delay seconds for livenessProbe	5
worker.livenessProbe.periodSeconds	Period seconds for livenessProbe	30
worker.livenessProbe.timeoutSeconds	Timeout seconds for livenessProbe	5
worker.livenessProbe.failureThreshold	Failure threshold for livenessProbe	5
worker.livenessProbe.successThreshold	Success threshold for livenessProbe	1
worker.readinessProbe.enabled	Enable readinessProbe on Worker nodes	true
worker.readinessProbe.initialDelaySeconds	Initial delay seconds for readinessProbe	5
worker.readinessProbe.periodSeconds	Period seconds for readinessProbe	30
worker.readinessProbe.timeoutSeconds	Timeout seconds for readinessProbe	5
worker.readinessProbe.failureThreshold	Failure threshold for readinessProbe	5
worker.readinessProbe.successThreshold	Success threshold for readinessProbe	1
worker.startupProbe.enabled	Enable startupProbe on Worker containers	false
worker.startupProbe.initialDelaySeconds	Initial delay seconds for startupProbe	5
worker.startupProbe.periodSeconds	Period seconds for startupProbe	30
worker.startupProbe.timeoutSeconds	Timeout seconds for startupProbe	5
worker.startupProbe.failureThreshold	Failure threshold for startupProbe	5
worker.startupProbe.successThreshold	Success threshold for startupProbe	1
worker.customLivenessProbe	Custom livenessProbe that overrides the default one	{}
worker.customReadinessProbe	Custom readinessProbe that overrides the default one	{}
worker.customStartupProbe	Custom startupProbe that overrides the default one	{}
worker.resourcesPreset	Set container resources according to one common preset (allowed values: none, nano, micro, small, medium, large, xlarge, 2xlarge). This is ignored if worker.resources is set (worker.resources is recommended for production).	medium
worker.resources	Set container requests and limits for different resources like CPU or memory (essential for production workloads)	{}
worker.podSecurityContext.enabled	Enabled Worker pods’ Security Context	true
worker.podSecurityContext.fsGroupChangePolicy	Set filesystem group change policy	Always
worker.podSecurityContext.sysctls	Set kernel settings using the sysctl interface	[]
worker.podSecurityContext.supplementalGroups	Set filesystem extra groups	[]
worker.podSecurityContext.fsGroup	Set Worker pod’s Security Context fsGroup	1001
worker.containerSecurityContext.enabled	Enabled Worker containers’ Security Context	true
worker.containerSecurityContext.seLinuxOptions	Set SELinux options in container	{}
worker.containerSecurityContext.runAsUser	Set Worker containers’ Security Context runAsUser	1001
worker.containerSecurityContext.runAsGroup	Set Worker containers’ Security Context runAsGroup	1001
worker.containerSecurityContext.runAsNonRoot	Set Worker containers’ Security Context runAsNonRoot	true
worker.containerSecurityContext.readOnlyRootFilesystem	Set Worker containers’ Security Context runAsNonRoot	true
worker.containerSecurityContext.allowPrivilegeEscalation	Set Worker container’s privilege escalation	false
worker.containerSecurityContext.capabilities.drop	Set Worker container’s Security Context runAsNonRoot	["ALL"]
worker.containerSecurityContext.seccompProfile.type	Set Worker container’s Security Context seccomp profile	RuntimeDefault
worker.containerSecurityContext.privileged	Set Worker container’s Security Context privileged	false
worker.lifecycleHooks	for the client container(s) to automate configuration before or after startup	{}
worker.runtimeClassName	Name of the runtime class to be used by pod(s)	""
worker.automountServiceAccountToken	Mount Service Account token in pod	false
worker.hostAliases	client pods host aliases	[]
worker.labels	Labels for the worker deployment	{}
worker.annotations	Annotations for the worker deployment	{}
worker.podLabels	Extra labels for client pods	{}
worker.podAnnotations	Annotations for client pods	{}
worker.podAffinityPreset	Pod affinity preset. Ignored if client.affinity is set. Allowed values: soft or hard	""
worker.podAntiAffinityPreset	Pod anti-affinity preset. Ignored if client.affinity is set. Allowed values: soft or hard	soft
worker.nodeAffinityPreset.type	Node affinity preset type. Ignored if client.affinity is set. Allowed values: soft or hard	""
worker.nodeAffinityPreset.key	Node label key to match. Ignored if client.affinity is set	""
worker.nodeAffinityPreset.values	Node label values to match. Ignored if client.affinity is set	[]
worker.affinity	Affinity for Worker pods assignment	{}
worker.nodeSelector	Node labels for Worker pods assignment	{}
worker.tolerations	Tolerations for Worker pods assignment	[]
worker.topologySpreadConstraints	Topology Spread Constraints for pod assignment spread across your cluster among failure-domains	[]
worker.podManagementPolicy	Statefulset Pod management policy, it needs to be Parallel to be able to complete the cluster join	Parallel
worker.priorityClassName	Worker pods’ priorityClassName	""
worker.schedulerName	Kubernetes pod scheduler registry	""
worker.updateStrategy.type	Worker statefulset strategy type	RollingUpdate
worker.updateStrategy.rollingUpdate	Worker statefulset rolling update configuration parameters	{}
worker.extraVolumes	Optionally specify extra list of additional volumes for the Worker pod(s)	[]
worker.extraVolumeMounts	Optionally specify extra list of additional volumeMounts for the Worker container(s)	[]
worker.sidecars	Add additional sidecar containers to the Worker pod(s)	[]
worker.enableDefaultInitContainers	Deploy default init containers	true
worker.initContainers	Add additional init containers to the Worker pod(s)	[]
worker.headlessServiceAnnotations	Annotations for the headless service	{}

Worker Traffic Exposure Parameters

Name	Description	Value
worker.externalAccess.enabled	Create a service per worker node	false
worker.externalAccess.service.type	Worker service type	ClusterIP
worker.externalAccess.service.loadBalancerIPs	Array of load balancer IPs for each Kafka broker. Length must be the same as replicaCount	[]
worker.externalAccess.service.externalIPs	Use distinct service host IPs to configure Kafka external listener when service type is NodePort. Length must be the same as replicaCount	[]
worker.externalAccess.service.loadBalancerAnnotations	Array of load balancer annotations for each Kafka broker. Length must be the same as replicaCount	[]
worker.externalAccess.service.publishNotReadyAddresses	Indicates that any agent which deals with endpoints for this Service should disregard any indications of ready/not-ready	false
worker.externalAccess.service.ports.ssh	Worker GRPC service port	22
worker.externalAccess.service.nodePorts	Array of node ports used for each Kafka broker. Length must be the same as replicaCount	[]
worker.externalAccess.service.sessionAffinityConfig	Additional settings for the sessionAffinity	{}
worker.externalAccess.service.sessionAffinity	Control where client requests go, to the same pod or round-robin	None
worker.externalAccess.service.loadBalancerSourceRanges	Worker service Load Balancer sources	[]
worker.externalAccess.service.externalTrafficPolicy	Worker service external traffic policy	Cluster
worker.externalAccess.service.labels	Additional custom labels for Worker service	{}
worker.externalAccess.service.annotations	Additional custom annotations for Worker service	{}
worker.externalAccess.service.extraPorts	Extra ports to expose in the Worker service	[]
worker.serviceAccount.create	Enable creation of ServiceAccount for Data Coordinator pods	true
worker.serviceAccount.name	The name of the ServiceAccount to use	""
worker.serviceAccount.automountServiceAccountToken	Allows auto mount of ServiceAccountToken on the serviceAccount created	false
worker.serviceAccount.annotations	Additional custom annotations for the ServiceAccount	{}
worker.networkPolicy.enabled	Enable creation of NetworkPolicy resources	true
worker.networkPolicy.allowExternal	The Policy model to apply	true
worker.networkPolicy.allowExternalEgress	Allow the pod to access any range of port and all destinations.	true
worker.networkPolicy.extraIngress	Add extra ingress rules to the NetworkPolicy	[]
worker.networkPolicy.extraEgress	Add extra ingress rules to the NetworkPolicy	[]
worker.networkPolicy.ingressNSMatchLabels	Labels to match to allow traffic from other namespaces	{}
worker.networkPolicy.ingressNSPodMatchLabels	Pod labels to match to allow traffic from other namespaces	{}

Deepspeed Worker persistence paramaters

Name	Description	Value
worker.persistence.enabled	Use a PVC to persist data	false
worker.persistence.storageClass	discourse & sidekiq data Persistent Volume Storage Class	""
worker.persistence.existingClaim	Use a existing PVC which must be created manually before bound	""
worker.persistence.mountPath	Path to mount the volume at	/bitnami/deepspeed/data
worker.persistence.accessModes	Persistent Volume Access Mode	["ReadWriteOnce"]
worker.persistence.selector	Selector to match an existing Persistent Volume for the worker data PVC	{}
worker.persistence.dataSource	Custom PVC data source	{}
worker.persistence.size	Size of data volume	8Gi
worker.persistence.labels	Persistent Volume labels	{}
worker.persistence.annotations	Persistent Volume annotations	{}
worker.pdb.create	Enable/disable a Pod Disruption Budget creation	true
worker.pdb.minAvailable	Minimum number/percentage of pods that should remain scheduled	""
worker.pdb.maxUnavailable	Maximum number/percentage of pods that may be made unavailable. Defaults to 1 if both worker.pdb.minAvailable and worker.pdb.maxUnavailable are empty.	""
gitImage.registry	Git image registry	REGISTRY_NAME
gitImage.repository	Git image repository	REPOSITORY_NAME/git
gitImage.digest	Git image digest in the way sha256:aa…. Please note this parameter, if set, will override the tag	""
gitImage.pullPolicy	Git image pull policy	IfNotPresent
gitImage.pullSecrets	Specify docker-registry secret names as an array	[]
volumePermissions.enabled	Enable init container that changes volume permissions in the data directory	false
volumePermissions.image.registry	Init container volume-permissions image registry	REGISTRY_NAME
volumePermissions.image.repository	Init container volume-permissions image repository	REPOSITORY_NAME/os-shell
volumePermissions.image.digest	Init container volume-permissions image digest in the way sha256:aa…. Please note this parameter, if set, will override the tag	""
volumePermissions.image.pullPolicy	Init container volume-permissions image pull policy	IfNotPresent
volumePermissions.image.pullSecrets	Specify docker-registry secret names as an array	[]
volumePermissions.resourcesPreset	Set container resources according to one common preset (allowed values: none, nano, micro, small, medium, large, xlarge, 2xlarge). This is ignored if volumePermissions.resources is set (volumePermissions.resources is recommended for production).	nano
volumePermissions.resources	Set container requests and limits for different resources like CPU or memory (essential for production workloads)	{}

Specify each parameter using the --set key=value[,key=value] argument to helm install. For example,

helm install my-release \
  --set worker.replicaCount=4 \
    oci://REGISTRY_NAME/REPOSITORY_NAME/deepspeed

Note: You need to substitute the placeholders REGISTRY_NAME and REPOSITORY_NAME with a reference to your Helm chart registry and repository. For example, in the case of Bitnami, you need to use REGISTRY_NAME=registry-1.docker.io and REPOSITORY_NAME=bitnamicharts.

The above command create 4 pods for DeepSpeed workers.

Alternatively, a YAML file that specifies the values for the parameters can be provided while installing the chart. For example,

helm install my-release -f values.yaml oci://REGISTRY_NAME/REPOSITORY_NAME/deepspeed

Note: You need to substitute the placeholders REGISTRY_NAME and REPOSITORY_NAME with a reference to your Helm chart registry and repository. For example, in the case of Bitnami, you need to use REGISTRY_NAME=registry-1.docker.io and REPOSITORY_NAME=bitnamicharts. Tip: You can use the default values.yaml

Troubleshooting

Find more information about how to deal with common errors related to Bitnami’s Helm charts in this troubleshooting guide.

Upgrading

To 2.0.0

This major bump changes the following security defaults:

• resourcesPreset is changed from none to the minimum size working in our test suites (NOTE: resourcesPreset is not meant for production usage, but resources adapted to your use case).
• global.compatibility.openshift.adaptSecurityContext is changed from disabled to auto.

This could potentially break any customization or init scripts used in your deployment. If this is the case, change the default values to the previous ones.

License

Copyright © 2024 Broadcom. The term “Broadcom” refers to Broadcom Inc. and/or its subsidiaries.

Licensed under the Apache License, Version 2.0 (the “License”); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an “AS IS” BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.