qlinks package#

Subpackages#

Submodules#

qlinks.distributed module#

Generic distributed task execution utilities.

This module intentionally does not depend on model, eigensolver, or caging objects. It provides a small map-style abstraction that can be reused by caging searches, parameter sweeps, and future workflows.

class qlinks.distributed.DistributedConfig(backend='serial', preserve_order=True, show_progress=True, progress_description='Completed jobs', failure_mode='raise', ray_init_kwargs=None, ray_shutdown=False, num_cpus_per_task=None, num_gpus_per_task=None)[source]#

Bases: object

Configuration for generic distributed task execution.

Parameters:

backend (Literal['serial', 'ray'])
preserve_order (bool)
show_progress (bool)
progress_description (str)
failure_mode (Literal['raise', 'return'])
ray_init_kwargs (dict | None)
ray_shutdown (bool)
num_cpus_per_task (float | None)
num_gpus_per_task (float | None)

backend: Literal['serial', 'ray']#

preserve_order: bool#

show_progress: bool#

progress_description: str#

failure_mode: Literal['raise', 'return']#

ray_init_kwargs: dict | None#

ray_shutdown: bool#

num_cpus_per_task: float | None#

num_gpus_per_task: float | None#

__init__(backend='serial', preserve_order=True, show_progress=True, progress_description='Completed jobs', failure_mode='raise', ray_init_kwargs=None, ray_shutdown=False, num_cpus_per_task=None, num_gpus_per_task=None)#

Parameters:

backend (Literal['serial', 'ray'])
preserve_order (bool)
show_progress (bool)
progress_description (str)
failure_mode (Literal['raise', 'return'])
ray_init_kwargs (dict | None)
ray_shutdown (bool)
num_cpus_per_task (float | None)
num_gpus_per_task (float | None)

Return type:

None

class qlinks.distributed.TaskFailure(task_index, task_input, exception_type, message)[source]#

Bases: object

Container returned when a task fails and failure_mode=’return’.

Parameters:

task_index (int)
task_input (object)
exception_type (str)
message (str)

task_index: int#

task_input: object#

exception_type: str#

message: str#

__init__(task_index, task_input, exception_type, message)#

Parameters:

task_index (int)
task_input (object)
exception_type (str)
message (str)

Return type:

None

qlinks.distributed.map_tasks(function, inputs, *, config=None)[source]#

Map a function over inputs using the configured execution backend.

Parameters:

function (Callable[[InputType], OutputType])
inputs (Sequence[InputType])
config (DistributedConfig | None)

Return type:

list[OutputType] | list[OutputType | TaskFailure]

qlinks.distributed.map_tasks_serial(function, inputs, *, preserve_order=True, show_progress=True, progress_description='Completed jobs', failure_mode='raise')[source]#

Map tasks serially.

preserve_order is accepted for API compatibility. Serial execution naturally preserves input order.

Parameters:

function (Callable[[InputType], OutputType])
inputs (Sequence[InputType])
preserve_order (bool)
show_progress (bool)
progress_description (str)
failure_mode (Literal['raise', 'return'])

Return type:

list[OutputType] | list[OutputType | TaskFailure]

qlinks.distributed.map_tasks_ray(function, inputs, *, preserve_order=True, show_progress=True, progress_description='Completed jobs', failure_mode='raise', ray_init_kwargs=None, ray_shutdown=False, num_cpus_per_task=None, num_gpus_per_task=None)[source]#

Map tasks with Ray.

Parameters:

function (Callable[[InputType], OutputType])
inputs (Sequence[InputType])
preserve_order (bool)
show_progress (bool)
progress_description (str)
failure_mode (Literal['raise', 'return'])
ray_init_kwargs (dict | None)
ray_shutdown (bool)
num_cpus_per_task (float | None)
num_gpus_per_task (float | None)

Return type:

list[OutputType] | list[OutputType | TaskFailure]

qlinks.distributed.ray_task_wrapper(task_index, task_input, function, failure_mode)[source]#

Run one task and return its index together with the result.

Parameters:

task_index (int)
task_input (InputType)
function (Callable[[InputType], OutputType])
failure_mode (Literal['raise', 'return'])

Return type:

tuple[int, OutputType | TaskFailure]

qlinks.distributed.map_on_ray(func, params)[source]#

Backward-compatible Ray map.

Results preserve the historical behavior and are returned in completion order, not input order.

Parameters:

func (Callable[[InputType], OutputType])
params (Sequence[InputType])

Return type:

list[OutputType]

qlinks package

Contents

qlinks package#

Subpackages#

Submodules#

qlinks.distributed module#

Module contents#