Machine Learning Exchange (MLX)
0 likes•344 views
Machine Learning Exchange (MLX) is a catalog and execution engine for AI assets including pipelines, models, datasets and notebooks. It allows users to upload, register, execute and deploy these assets. MLX generates sample pipeline code and uses Kubeflow Pipelines powered by Tekton as its pipelines engine. It integrates with services like KFServing for model serving, Dataset Lifecycle Framework for data management, and MAX/DAX for pre-registered datasets and models. MLX provides APIs, UI and SDK to interact with these AI assets.
![Pluggable Components
Watson
Studio WML
Open
Scale
Kubeflow
Training
Seldon AIF360 ART KATIB KFSERVING
…
…
TASK
STEP
POD
STEP
TASK
STEP STEP
POD
Container Container Container Container
TEKTON
KFP API Server
Components
Pipelines
Object
Store
KFP UI
Relational
DB
COMPILE
KFP SDK
Intermediate
Representation [IR]
Pipelines - default integration with Kubeflow Pipelines and Tekton](https://image.slidesharecdn.com/mlx-up-210517204203/85/Machine-Learning-Exchange-MLX-12-320.jpg)
Machine Learning Exchange (MLX)
- 1. Machine Learning Exchange (MLX) Animesh Singh, Christian Kadner, Tommy Li, Andrew Butler
- 2. Machine Learning Exchange (MLX) : Data and AI Assets Catalog and Execution Engine
- 3. Machine Learning Exchange (MLX) – Data and AI Assets Catalog and Execution Engine – Upload, register, execute, and deploy -AI pipelines and Components -Models -Datasets -Notebooks – Automated sample pipeline code generation to train, validate, serve your registered models, datasets and notebooks – Pipelines Engine powered by Kubeflow Pipelines on Tekton, core of Watson Pipelines – Serving engine by KFServing (Next gen base for WML) , Datasets Management by Dataset Lifecycle Framework – Preregistered Datasets from Data Asset Exchange (DAX) and Models from Model Asset Exchange (MAX) – Model Metadata schema aligned with MLSpec MLX API Server Components Pipelines Notebooks Object Store Kubeflow Pipelines on Tekton MLX UI SDK Relational DB Models Datasets Datashim KFServing Notebook Component - Elyra MAX/DAX
- 4. View, download, and execute Pipelines 4
- 5. View, download, and execute Pipeline Components 5
- 6. Library of prepackaged models. Register your own models, run with Pipelines 6
- 7. View, modify, and monitor deployed models 7
- 8. Register and deploy Datasets (as sharable Volumes for other assets) 8
- 9. Library of prepackaged notebooks. Register your own notebooks
- 10. 10 Run Notebooks using Pipelines
- 11. Integration
- 12. Pluggable Components Watson Studio WML Open Scale Kubeflow Training Seldon AIF360 ART KATIB KFSERVING … … TASK STEP POD STEP TASK STEP STEP POD Container Container Container Container TEKTON KFP API Server Components Pipelines Object Store KFP UI Relational DB COMPILE KFP SDK Intermediate Representation [IR] Pipelines - default integration with Kubeflow Pipelines and Tekton
- 13. Datasets - default integration with Datashim
- 14. Models – default integration with KFServing Kubernetes Compute cluster GPU, TPU ,CPU Model Assets. Istio Knative KFServing PRE-PROCESS PREDICT POST-PROCESS EXPLAIN
- 15. Notebooks – default integration with Elyra Notebook Component