Execution frameworks
Choose an execution framework based on what you need to run, how it should run, and which dependencies it requires. DigitalHub provides execution frameworks for Python jobs, containers, model serving, workflow orchestration, data transformation, federated learning, and service pipelines.
Description
Execute user-defined Python handlers for jobs, model training, and services.
Use cases
- Batch jobs and model training.
- Services and APIs.
Description
Run existing containerized applications with custom dependencies.
Use cases
- Existing containerized applications.
- Workloads with custom system dependencies.
- Remote jobs and services on Kubernetes.
Description
Deploy supported machine learning models as scalable inference services.
Use cases
- REST inference endpoints.
- scikit-learn, MLflow, and other supported models.
- vLLM text and speech serving.
Description
Build and execute multi-step workflows with Hera.
Use cases
- DAG and Steps pipelines.
- Conditional or parallel workflow execution.
- Workflow definitions compiled for remote runs.
Description
Build privacy-preserving federated-learning workloads.
Use cases
- Federated learning simulations.
- Flower clients and servers.
- Training across distributed datasets.
Description
Transform tabular data with SQL-based workflows.
Use cases
- SQL transformations on tabular data.
- Local or remote DBT runs.
- PostgreSQL-backed data workflows.
Description
Deploy synchronous or asynchronous pipelines that orchestrate services and streaming data.
Use cases
- Synchronous or asynchronous service pipelines.
- AI service processing chains.
- Streaming inputs and outputs.
Description
Execute Python handlers for request/response processing using EnvoyProxy ExtProc specifications.
Use cases
- Request and response processing.
- Preprocessing and postprocessing traffic.
- Guardrails around service requests and responses.
Description
Execute Python handlers for inference-oriented services with explicit tensor schemas.
Use cases
- Inference endpoints.
- Explicit input and output tensor schemas.
- Model inference services.
Description
Build, compile, and serve machine learning models with Apache TVM.
Use cases
- Converting ONNX and TFLite models to Relax IR.
- Compiling models for a target architecture.
- Deploying compiled models as inference services.
Description
Run Hydra Python Job applications.
Use cases
- Complex Python jobs with multiple confgurations.
- Hyper Parameter Optimization with Optuna.
Next steps
Once you've selected an execution framework, follow these steps:
- Read the overview for your chosen execution framework.
- Select an action from its Actions section.
- Follow the Action, Create, Run and Use pages for the action's behavior, parameters and results.
- Check the examples for complete usage patterns.
- Explore the entity documentation for the complete API reference.
Still unsure?
Start with Python for general-purpose workloads. Choose Container for complex dependencies, Hera for coordinated steps, ModelServe for model inference, DBT for SQL transformations, or Flower for federated learning.