GS-aiflow
GS-aiflow AI/ML Workflow Management Framework
GS-aiflow is the workflow management Framework for machine learning operations - pipelines, training and inferences. When workflows are defined, they become more maintainable, versionable, testable, and collaborative. With GS-aiflow you can use workflows as directed acyclic graphs (DAGs) of tasks. The GS-aiflow scheduler executes your tasks on an array of workers while following the specified dependencies.
Aiflow offers a set of lightweight environments that can be used with any existing machine learning application or library (TensorFlow, PyTorch, Keras, ONNX etc), wherever you currently run ML/DL code (e.g. in notebooks, standalone applications).
Requirements
| Main version (dev) | Stable version (1.5) | |
|---|---|---|
| Python | 3.7, 3.8, 3.9 | 3.7, 3.8, 3.9 |
| Docekr | 18.09.x, 20.10.x | 18.09.x, 20.10.x |
| Kubernetes | 1.20, 1.19 | 1.22 |
| NVIDIA Docker | 20.10.17 | 20.10.17 |
Installation
System Architecture
- redis: message broker in GS-aiflow
- airflow: backend workflow framework in GS-aiflow
Features
- manage workflow with DAG
- automate configuration and management of tasks
- work with Kubernetes and Dockers
- optimize and accelerate ML/DL inferencing and training for fastest responce time
User Interface
Task view
Graph view
Contributing
If you're interested in being a contributor and want to get involved in developing the GEdge Platform code, please see DOCUMENTATIONs for details on submitting patches and the contribution workflow.
Community
We have a project site for the GEdge Platform. If you're interested in being a contributor and want to get involved in developing the Cloud Edge Platform code, please visit GEdge Plaform Project site
License
GEdge Platform is under the Apache 2.0 license. See the LICENSE file for details.