The task scheduler with customized methods on the python-side. More...

#include <task_scheduler.h>

Inheritance diagram for tvm::meta_schedule::PyTaskSchedulerNode:

Collaboration diagram for tvm::meta_schedule::PyTaskSchedulerNode:

Public Types
using	FNextTaskId = ffi::TypedFunction< int()>
	The function type of `NextTaskId` method. More...

using	FJoinRunningTask = ffi::TypedFunction< Array< RunnerResult >(int)>
	The function type of `JoinRunningTask` method. More...

using	FTune = ffi::TypedFunction< void(Array< TuneContext > tasks, Array< FloatImm > task_weights, int max_trials_global, int max_trials_per_task, int num_trials_per_iter, Builder builder, Runner runner, Array< MeasureCallback > measure_callbacks, Optional< Database > database, Optional< CostModel > cost_model)>
	The function type of `Tune` method. More...

Public Member Functions
int	NextTaskId () final
	Fetch the next task id. More...

Array< RunnerResult >	JoinRunningTask (int task_id) final
	Wait until the task is finished. More...

void	Tune (Array< TuneContext > tasks, Array< FloatImm > task_weights, int max_trials_global, int max_trials_per_task, int num_trials_per_iter, Builder builder, Runner runner, Array< MeasureCallback > measure_callbacks, Optional< Database > database, Optional< CostModel > cost_model) final
	Jointly tune a given list of tasks. More...

	TVM_DECLARE_FINAL_OBJECT_INFO (PyTaskSchedulerNode, TaskSchedulerNode)

Public Member Functions inherited from tvm::meta_schedule::TaskSchedulerNode
virtual	~TaskSchedulerNode ()=default
	The default destructor. More...

void	TerminateTask (int task_id)
	Terminate a task. More...

void	TouchTask (int task_id)
	Touch the task and update its status. More...

void	PrintTuningStatistics ()
	Print out a human-readable format of the tuning statistics. More...

	TVM_DECLARE_BASE_OBJECT_INFO (TaskSchedulerNode, Object)

Static Public Member Functions
static void	RegisterReflection ()

Static Public Member Functions inherited from tvm::meta_schedule::TaskSchedulerNode
static void	RegisterReflection ()

Public Attributes
FNextTaskId	f_next_task_id
	The packed function to the `NextTaskId` function. More...

FJoinRunningTask	f_join_running_task
	The packed function to the `JoinRunningTask` function. More...

FTune	f_tune
	The packed function to the `Tune` function. More...

Public Attributes inherited from tvm::meta_schedule::TaskSchedulerNode
ffi::Function	logger
	The tuning task's logging function. More...

Array< TaskRecord >	tasks_
	Records for each task. More...

Array< MeasureCallback >	measure_callbacks_
	The list of measure callbacks of the scheduler. More...

Optional< Database >	database_
	The database used in tuning. More...

Optional< CostModel >	cost_model_
	The cost model used in tuning. More...

int	remaining_tasks_
	The number of remaining tasks to be tuned. More...

Static Public Attributes
static constexpr const char *	_type_key = "meta_schedule.PyTaskScheduler"

Static Public Attributes inherited from tvm::meta_schedule::TaskSchedulerNode
static constexpr const char *	_type_key = "meta_schedule.TaskScheduler"

Detailed Description

The task scheduler with customized methods on the python-side.

Member Typedef Documentation

◆ FJoinRunningTask

using tvm::meta_schedule::PyTaskSchedulerNode::FJoinRunningTask = ffi::TypedFunction<Array<RunnerResult>(int)>

The function type of JoinRunningTask method.

Parameters

task_id The task id to be joined.

◆ FNextTaskId

using tvm::meta_schedule::PyTaskSchedulerNode::FNextTaskId = ffi::TypedFunction<int()>

The function type of NextTaskId method.

Returns: The next task id.

◆ FTune

using tvm::meta_schedule::PyTaskSchedulerNode::FTune = ffi::TypedFunction<void(Array<TuneContext> tasks, Array<FloatImm> task_weights, int max_trials_global, int max_trials_per_task, int num_trials_per_iter, Builder builder, Runner runner, Array<MeasureCallback> measure_callbacks, Optional<Database> database, Optional<CostModel> cost_model)>

The function type of Tune method.

Member Function Documentation

◆ JoinRunningTask()

Array<RunnerResult> tvm::meta_schedule::PyTaskSchedulerNode::JoinRunningTask ( int task_id )

finalvirtual

Wait until the task is finished.

Parameters

task_id The task id to be joined.

Returns: The results from the runner.

Reimplemented from tvm::meta_schedule::TaskSchedulerNode.

◆ NextTaskId()

int tvm::meta_schedule::PyTaskSchedulerNode::NextTaskId ( )

finalvirtual

Fetch the next task id.

Returns: The next task id.

Implements tvm::meta_schedule::TaskSchedulerNode.

◆ RegisterReflection()

static void tvm::meta_schedule::PyTaskSchedulerNode::RegisterReflection ( )

inlinestatic

◆ Tune()

void tvm::meta_schedule::PyTaskSchedulerNode::Tune	(	Array< TuneContext >	tasks,
		Array< FloatImm >	task_weights,
		int	max_trials_global,
		int	max_trials_per_task,
		int	num_trials_per_iter,
		Builder	builder,
		Runner	runner,
		Array< MeasureCallback >	measure_callbacks,
		Optional< Database >	database,
		Optional< CostModel >	cost_model
	)

finalvirtual

Jointly tune a given list of tasks.

Parameters

tasks	The tasks to be tuned
task_weights	The weight of each task
max_trials_global	The maximum number of trials to be performed globally
max_trials_per_task	The maximum number of trials to be performed for each task
num_trials_per_iter	The number of trials to be performed in each iteration
builder	The MetaSchedule builder
runner	The MetaSchedule runner
measure_callbacks	The callbacks to be called after each measurement
database	The database used in tuning
cost_model	The cost model used in tuning

Reimplemented from tvm::meta_schedule::TaskSchedulerNode.

◆ TVM_DECLARE_FINAL_OBJECT_INFO()

tvm::meta_schedule::PyTaskSchedulerNode::TVM_DECLARE_FINAL_OBJECT_INFO	(	PyTaskSchedulerNode	,
		TaskSchedulerNode
	)

Member Data Documentation

◆ _type_key

constexpr const char* tvm::meta_schedule::PyTaskSchedulerNode::_type_key = "meta_schedule.PyTaskScheduler"

staticconstexpr

◆ f_join_running_task

FJoinRunningTask tvm::meta_schedule::PyTaskSchedulerNode::f_join_running_task

The packed function to the JoinRunningTask function.

◆ f_next_task_id

FNextTaskId tvm::meta_schedule::PyTaskSchedulerNode::f_next_task_id

The packed function to the NextTaskId function.

◆ f_tune

FTune tvm::meta_schedule::PyTaskSchedulerNode::f_tune

The packed function to the Tune function.

The documentation for this class was generated from the following file:

include/tvm/meta_schedule/task_scheduler.h

Public Types

Public Member Functions

Static Public Member Functions

Public Attributes

Static Public Attributes

Detailed Description

Member Typedef Documentation

◆ FJoinRunningTask

◆ FNextTaskId

◆ FTune

Member Function Documentation

◆ JoinRunningTask()

◆ NextTaskId()

◆ RegisterReflection()

◆ Tune()

◆ TVM_DECLARE_FINAL_OBJECT_INFO()

Member Data Documentation

◆ _type_key

◆ f_join_running_task

◆ f_next_task_id

◆ f_tune