Sequential Strategy

Chain operators to execute one after another.

See Also

• Operators Overview - All operator types
• Composite Operator - Operator composition
• Parallel Strategy - Parallel execution
• Pipeline Tutorial

---

datarax.operators.strategies.sequential

Sequential composition strategies.

logger module-attribute

logger = getLogger(__name__)

SequentialStrategy

Bases: CompositionStrategyImpl

Applies operators in sequence (chain).

apply

apply(operators: list[OperatorModule], context: StrategyContext) -> tuple[PyTree, PyTree, dict[str, Any]]

Apply operators sequentially, piping each output to the next.

Parameters:

Name	Type	Description	Default
operators	list[OperatorModule]	Ordered list of operators to chain.	required
context	StrategyContext	Execution context with input data, state, and RNG params.	required

Returns:

Type	Description
tuple[PyTree, PyTree, dict[str, Any]]	Tuple of (data, state, metadata) after all operators have run.

describe

describe() -> dict[str, Any]

Return a serializable description of this strategy.

ConditionalSequentialStrategy

ConditionalSequentialStrategy(conditions: Sequence[Callable[[PyTree], bool | Array]])

Bases: CompositionStrategyImpl

Applies operators sequentially with conditions.

Only applies operators where condition evaluates to True.

Parameters:

Name	Type	Description	Default
conditions	Sequence[Callable[[PyTree], bool | Array]]	List of callables that determine whether each operator is applied.	required

conditions instance-attribute

conditions = conditions

apply

apply(operators: list[OperatorModule], context: StrategyContext) -> tuple[PyTree, PyTree, dict[str, Any]]

Apply operators sequentially, skipping those whose condition is False.

Uses jax.lax.cond for vmap-compatible conditional execution.

Parameters:

Name	Type	Description	Default
operators	list[OperatorModule]	Operators to apply (must match length of conditions).	required
context	StrategyContext	Execution context with input data, state, and RNG params.	required

Returns:

Type	Description
tuple[PyTree, PyTree, dict[str, Any]]	Tuple of (data, state, metadata) after conditional execution.

Raises:

Type	Description
ValueError	If operator count doesn't match condition count.

describe

describe() -> dict[str, Any]

Return a serializable description of this strategy.