ack-ai-pipeline is a workflow orchestration engine in the cloud-native AI suite that provides job orchestration, experiment management, and model tracking to simplify AI production task management and improve model iteration efficiency. This topic describes ack-ai-pipeline basics, usage, and release notes. Built on Kubeflow Pipelines and optimized for stability, security, and operational simplicity, it lets you build an MLOps platform on ACK without maintaining the upstream open-source stack.
Capabilities
Job orchestration — Run distributed Argo Workflows on Kubernetes clusters as directed-acyclic graph (DAG) workflows. Define, manage, and reuse DAG workflows on demand.
Experiment management — Compare workflow runs across experiment parameters, such as model hyperparameters, to fine-tune jobs.
Model tracking — Track the input and output of each workflow step to troubleshoot and fine-tune models.
Supported clusters
ack-ai-pipeline runs on ACK Pro and ACK Edge Pro clusters. Minimum Kubernetes version: 1.18.
| Cluster type | Kubernetes version |
|---|---|
| ACK Pro cluster | 1.18 and later |
| ACK Edge Pro cluster | 1.18 and later |
Next steps
Release notes
May 2023
| Version | Description | Release date | Impact |
|---|---|---|---|
| 1.0.3 | Enhanced security. | 2023-05-11 | No impact on workloads |
April 2023
| Version | Description | Release date | Impact |
|---|---|---|---|
| 1.0.2 | Enhanced security. | 2023-04-17 | No impact on workloads |
March 2023
| Version | Description | Release date | Impact |
|---|---|---|---|
| 1.0.1 | Fixed OSS bucket name conflicts. | 2023-03-14 | No impact on workloads |
July 2022
| Version | Description | Release date | Impact |
|---|---|---|---|
| 1.0.0 | Initial release. | 2022-07-04 | No impact on workloads |