fastvideo.workflow.workflow_base

fastvideo.workflow.workflow_base#

Module Contents#

Classes#

WorkflowBase

Abstract base class for defining video processing workflows.

Data#

logger

API#

class fastvideo.workflow.workflow_base.WorkflowBase(fastvideo_args: fastvideo.fastvideo_args.FastVideoArgs)[source]#

Bases: abc.ABC

Abstract base class for defining video processing workflows.

A workflow serves as the top-level orchestrator that coordinates multiple pipelines and components to accomplish a specific video processing task. The workflow pattern provides several key benefits:

  1. Separation of Concerns: Workflows separate high-level orchestration logic from low-level processing implementations in pipelines.

  2. Modularity: Different workflows can be created for different execution modes (preprocess, inference, etc.) while sharing common pipeline components.

  3. Configuration Management: Workflows manage the configuration and initialization of multiple related pipelines and components in a centralized manner.

  4. Environment Setup: Workflows handle system-level setup and resource allocation before pipeline execution begins.

  5. Lifecycle Management: Workflows control the complete lifecycle from initialization through execution to cleanup.

The workflow acts as a factory and coordinator, creating the appropriate pipelines based on configuration, setting up the execution environment, and orchestrating the overall processing flow.

Initialization

Initialize the workflow with configuration arguments.

Parameters:

fastvideo_args – Configuration object containing all parameters needed for workflow and pipeline setup.

add_component(component_name: str, component: Any) → None[source]#

Register a component instance with the workflow.

Components are auxiliary objects that may be shared across pipelines or used for workflow-level functionality (e.g., databases, caches, external services).

Parameters:

  • component_name – Unique identifier for the component.

  • component – The component instance to register.

add_pipeline_config(pipeline_name: str, pipeline_config: tuple[fastvideo.pipelines.pipeline_registry.PipelineType, fastvideo.fastvideo_args.FastVideoArgs]) → None[source]#

Register a pipeline configuration for later instantiation.

Parameters:

  • pipeline_name – Unique identifier for the pipeline.

  • pipeline_config – Tuple containing the pipeline type and configuration arguments.

get_component(component_name: str) → Any[source]#

Retrieve a registered component by name.

Parameters:

component_name – The name of the component to retrieve.

Returns:

The component instance.

classmethod get_workflow_cls(fastvideo_args: fastvideo.fastvideo_args.FastVideoArgs) → Optional[fastvideo.workflow.workflow_base.WorkflowBase][source]#

Factory method to get the appropriate workflow class based on execution mode.

This method acts as a workflow factory, returning the appropriate workflow class implementation based on the specified execution mode in the configuration arguments.

Parameters:

fastvideo_args – Configuration object containing the execution mode and other parameters.

Returns:

The appropriate workflow class for the specified execution mode, or None if no workflow is available for the given mode.

load_pipelines() → None[source]#

Create and initialize all registered pipelines.

This method instantiates pipeline objects from their configurations and makes them available as both dictionary entries and instance attributes for convenient access.

abstract prepare_system_environment() → None[source]#

Prepare the system environment for workflow execution.

Subclasses must implement this method to handle any system-level setup required before pipeline execution (e.g., GPU initialization, temporary directories, resource allocation).

abstract register_components() → None[source]#

Register workflow-specific components.

Subclasses must implement this method to register any components needed for their specific workflow (e.g., databases, external APIs, shared resources).

abstract register_pipelines() → None[source]#

Register workflow-specific pipelines.

Subclasses must implement this method to define which pipelines are needed for their specific workflow and how they should be configured.

abstract run()[source]#

Execute the main workflow logic.

Subclasses must implement this method to define the specific execution flow for their workflow, coordinating the registered pipelines and components to accomplish the desired task.

fastvideo.workflow.workflow_base.logger[source]#

‘init_logger(…)’

Contents