View Code? Open in Web Editor
NEW
Baklava is the build and packaging system for ML models. Baklava leverages the python standard "setuptools" packaging system, and extends it to build docker containers that run Machine learning models. These containers are compatible with SageMaker, and in future, they will be compatible with Kubeflow.
License: Apache License 2.0
Python 96.58%
Dockerfile 3.42%
baklava's People
baklava's Issues
๐ Feature Proposal
A clear and concise description of what the feature is.
Motivation
Please outline the motivation for the proposal.
Similar to .git, .terraform, this change make the build process hidden and can also be merged with mlctl temp files when creating a new job.
Example
Please provide an example for how this feature would be used.
๐ Feature Proposal
Integrate with Github Actions and enable a continuous build and deployment of the baklava libraries to pypi.
๐ Feature Proposal
When Baklava builds a serving
container using the predict
function, make the serving container compatible with the KFServing V2 API specification
Motivation
This capability will enable containers produced by Baklava to be used directly on any KFServing compatible server, such as the Triton Inference Server
๐ Feature Proposal
Add support in Baklava to produce containers that run on GCP for both training and inference
Motivation
This capability will enable Baklava to be compatible with multiple clouds (currently supports only AWS SageMaker)
Example
python setup.py train --target=gcp
๐ Feature Proposal
A clear and concise description of what the feature is.
Motivation
Please outline the motivation for the proposal.
Example
Please provide an example for how this feature would be used.