No project description provided
Project description
Soopervisor
Soopervisor runs Ploomber pipelines for batch processing (large-scale training or batch serving) or online inference.
pip install soopervisor
Check out the documentation to learn more.
Compatible with Python 3.6 and higher.
Supported platforms
Batch serving and large-scale training:
Online inference:
From notebook to a production pipeline
We also have an example that shows how to use our ecosystem of tools to go from a monolithic notebook to a pipeline deployed in Kubernetes.
Usage
Say that you want to train multiple models in a Kubernetes cluster, you may create a new target environment to execute your pipeline using Argo Workflows:
soopervisor add training --backend argo-workflows
After filling in some basic configuration settings, export the pipeline with:
soopervisor export training
Depending on the selected backend (Argo, Airflow, AWS Batch, or AWS Lambda), configuration details will change, but the API remains the same: soopervisor add, then soopervisor export. CHANGELOG =========
0.7 (2022-01-31)
Improves CLI documentation
Adds --git-ignore to documentation
Various documentation improvements
Changes short version of --until-build to -u
Adds “Task Communication” user guide
Display warnings if passing CLI options that do not apply to SLURM
SLURM exporter raises error if sbatch isn’t installed
Showing a warning if source dist is >5MB (#53)
soopervisor add adds a default exclude value by extracting product paths from pipeline.yaml
Copying user settings when generating the Docker image
Experimental Kubeflow integration
Airflow integration allows to choose between BashOperator, KubernetesPodOperator, and DockerOperator using --preset
Many modules and test cases re-written for better code quality and maintainability
0.6.1 (2022-01-16)
Fixes output message after exporting to Argo
Adds flag to source.copy to ignore git
soopervisor export raises and error if git isn’t tracking any files
Adds --git-ignore to soopervisor export
0.6 (2022-01-04)
Adds support for SLURM
AirflowExporter uses KubernetesPodOperator by default (#33)
Simplified Airflow and Argo/k8s tutorials
0.5.2 (2022-01-02)
Clearer error message when pending git commit
Clearer error message when the user does not change docker repository default value (#29)
Argo spec sets imagePullPolicy to Never if repository is null
Documents Kubernetes/Argo configuration schema
General documentation improvements
0.5.1 (2021-07-26)
Better error message when lock files do not exist
Documentation note on when using shared disks (must pass --skip-tests)
Adds build as a dependency
Check for lock files before creating soopervisor.yaml
Fixes docker.build issue that caused a pipeline.yaml to be used even when the environment required one with another name
Fixes error that caused AWS batch args to be passed as a single str
0.5 (2021-07-09)
load_tasks tries to initialize a spec matching the target env name (e.g., training target looks for pipeline.train.yaml
Compatibility fixes with Ploomber (requires >=0.12.1)
Fixes an error that caused Dockerfile to include a line to install project as a package even if there was no setup.py file
0.4.2 (2021-06-04)
Adds exclude to ignore files/directories from docker image
Adds user guide section to documentation
0.4.1 (2021-05-31)
Adds --mode option to soopervisor export
Batch export stops if there are no tasks to execute
Adds --skip-tests option to skip tests before submitting
0.4 (2021-05-22)
Important: Soopervisor was re-written. Some modules were deprecated and the API changed. This new architecture allows us to greatly simplify user experience and easily incorporate more platforms in the future.
New CLI
New documentation
New (simplified) soopervisor.yaml configuration schema
Support for non-packaged projects (i.e., the ones without a setup.py file)
Support for AWS Batch
Support for AWS Lambda
Argo Workflows integration builds a docker image
Airflow integration produces a DAG with DockerOperator tasks
Deprecates build module
Deprecates script module
Deprecates Box integration
0.3.4 (2021-04-18)
Export projects compatible with ploomber.OnlineModel to AWS Lambda
Allow initialization from empty soopervisor.yaml
0.3.3 (2021-03-07)
Support to pass extra cli args to ploomber task (via args in soopervisor.yaml) when running in Argo and Airflow
0.3.2 (2021-02-13)
Adds --root arg to soopervisor export-airflow to select an alternative project’s root
Determines default entry point using Ploomber’s API to allow automated discovery of pipeline.yaml in package layouts (e.g. src/package/pipeline.yaml)
0.3.1 (2021-02-11)
Changes to the Airflow generated DAG
Fixes a bug when initializing configuration from projects whose root is not the current directory
0.3 (2021-01-24)
env.airflow.yaml optional when exporting to Airflow (#17)
Validating exported argo YAML spec
Output argo YAML spec displays script in literal mode to make it readable
Fixed extra whitespace in generated script
Refactors ArgoMountedVolume to provide flexibility for different types of k8s volumes
Adds section in the documentation to run examples using minikube
Adds a few echo statements to generated script to provide better status feedback
0.2.2 (2020-11-21)
Adds ability to skip dag loading during project validation
Box uploader imported only if needed
Exposes option to skip dag loading from the CLI
0.2.1 (2020-11-20)
Adds Airflow DAG export
Adds Argo/Kubernetes DAG export
Support for uploading products to Box
0.2 (2020-10-15)
Adds DockerExecutor
Products are saved in a folder with the name of the current commit by default
Conda environments are created locally in a .soopervisor/ folder
Conda environments are cached by default
Ability to customize arguments to ploomber build
0.1 (2020-08-09)
First release
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for soopervisor-0.7-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | f8a3e5022f5ee3a1166971c769ae23676b0ab9aa0fe30e1f83af97740720af22 |
|
MD5 | 808d0b8bef0dc1a7c629c7aa830023e8 |
|
BLAKE2b-256 | 8191dfbb1b30e4237529be2e4a7f05f0488f6da77f3708f1b038b302be87f359 |