learning

package
v0.3.2 Latest Latest
Warning

This package is not in the latest version of its module.

 Go to latest Published: Feb 21, 2026 License: MIT Imports: 8 Imported by: 0

Documentation

Index

Constants

View Source 
const (
	// DefaultMaxWorkers is the default number of concurrent observation workers.
	DefaultMaxWorkers = 100
)

Variables

This section is empty.

Functions

func ConvertImpact 

func ConvertImpact(impact *ResourceImpact) *decision.ResourceImpact

ConvertImpact converts learning.ResourceImpact to decision.ResourceImpact.

Types

type AllStats 

type AllStats struct {
	Tasks      map[string]*TaskStats `json:"tasks"`
	TotalTasks int64                 `json:"total_tasks"`
}

AllStats holds statistics for all task types.

type Engine 

type Engine struct {
	// contains filtered or unexported fields
}

Engine coordinates learning from task executions.

func NewEngine 

func NewEngine(model Model, aggregator *monitor.Aggregator, observationDelay time.Duration, logger *slog.Logger) *Engine

NewEngine creates a new learning engine.

func NewEngineWithWorkers 

func NewEngineWithWorkers(model Model, aggregator *monitor.Aggregator, observationDelay time.Duration, logger *slog.Logger, maxWorkers int) *Engine

NewEngineWithWorkers creates a new learning engine with custom worker limit.

func (*Engine) ActiveWorkers 

func (e *Engine) ActiveWorkers() int

ActiveWorkers returns the current number of active observation workers.

func (*Engine) GetStats 

func (e *Engine) GetStats() *AllStats

GetStats returns statistics for all observed tasks.

func (*Engine) GetTaskStats 

func (e *Engine) GetTaskStats(task string) *TaskStats

GetTaskStats returns statistics for a specific task.

func (*Engine) Model 

func (e *Engine) Model() Model

Model returns the underlying model.

func (*Engine) NotifyTaskStart 

func (e *Engine) NotifyTaskStart(task string, complexity int)

NotifyTaskStart records that a task has started. It captures a baseline of system state and schedules an observation.

func (*Engine) PendingCount 

func (e *Engine) PendingCount() int

PendingCount returns the number of tasks awaiting observation.

func (*Engine) Predict 

func (e *Engine) Predict(task string, complexity int) *ResourceImpact

Predict returns predicted resource impact for a task.

func (*Engine) Stop 

func (e *Engine) Stop()

Stop gracefully shuts down the engine, waiting for pending observations.

func (*Engine) StopWithTimeout 

func (e *Engine) StopWithTimeout(timeout time.Duration)

StopWithTimeout stops the engine with a timeout.

type Model 

type Model interface {
	// Name returns the model name.
	Name() string

	// Observe records an observation of task impact.
	Observe(task string, complexity int, impact *ResourceImpact)

	// Predict returns predicted resource impact for a task.
	Predict(task string, complexity int) *ResourceImpact

	// GetStats returns statistics for all observed tasks.
	GetStats() *AllStats

	// GetTaskStats returns statistics for a specific task.
	GetTaskStats(task string) *TaskStats

	// SetObserver sets a callback that will be called when stats change.
	SetObserver(observer StatsObserver)

	// LoadStats loads previously saved statistics into the model.
	LoadStats(stats *AllStats)
}

Model is the interface for learning models that predict resource impact.

type ModelAdapter 

type ModelAdapter struct {
	// contains filtered or unexported fields
}

ModelAdapter wraps decision/model.PredictionModel to implement the old learning.Model interface. This allows gradual migration to the new model system.

func NewModelAdapter 

func NewModelAdapter(m model.PredictionModel) *ModelAdapter

NewModelAdapter creates a new adapter wrapping a PredictionModel.

func (*ModelAdapter) GetStats 

func (a *ModelAdapter) GetStats() *AllStats

GetStats returns statistics for all tasks.

func (*ModelAdapter) GetTaskStats 

func (a *ModelAdapter) GetTaskStats(task string) *TaskStats

GetTaskStats returns statistics for a specific task.

func (*ModelAdapter) LoadStats 

func (a *ModelAdapter) LoadStats(stats *AllStats)

LoadStats loads previously saved statistics. Note: This is a no-op for the new model system - use model.Load() instead.

func (*ModelAdapter) Name 

func (a *ModelAdapter) Name() string

Name returns the model name.

func (*ModelAdapter) Observe 

func (a *ModelAdapter) Observe(task string, complexity int, impact *ResourceImpact)

Observe records an observation.

func (*ModelAdapter) Predict 

func (a *ModelAdapter) Predict(task string, complexity int) *ResourceImpact

Predict returns predicted resource impact.

func (*ModelAdapter) SetObserver 

func (a *ModelAdapter) SetObserver(observer StatsObserver)

SetObserver sets a callback for stats changes.

func (*ModelAdapter) Underlying 

func (a *ModelAdapter) Underlying() model.PredictionModel

Underlying returns the wrapped PredictionModel.

type MovingAverageModel 

type MovingAverageModel struct {
	// contains filtered or unexported fields
}

MovingAverageModel implements Model using exponential moving average.

func NewMovingAverageModel 

func NewMovingAverageModel(alpha float64) *MovingAverageModel

NewMovingAverageModel creates a new MovingAverageModel. Alpha is the smoothing factor: higher values give more weight to recent observations. Typical values are 0.1-0.3.

func (*MovingAverageModel) GetStats 

func (m *MovingAverageModel) GetStats() *AllStats

func (*MovingAverageModel) GetTaskStats 

func (m *MovingAverageModel) GetTaskStats(task string) *TaskStats

func (*MovingAverageModel) LoadStats 

func (m *MovingAverageModel) LoadStats(stats *AllStats)

func (*MovingAverageModel) Name 

func (m *MovingAverageModel) Name() string

func (*MovingAverageModel) Observe 

func (m *MovingAverageModel) Observe(task string, complexity int, impact *ResourceImpact)

func (*MovingAverageModel) Predict 

func (m *MovingAverageModel) Predict(task string, complexity int) *ResourceImpact

func (*MovingAverageModel) SetObserver 

func (m *MovingAverageModel) SetObserver(observer StatsObserver)

type ResourceImpact 

type ResourceImpact struct {
	CPUDelta    float64 `json:"cpu_delta"`
	MemoryDelta float64 `json:"memory_delta"`
	GPUDelta    float64 `json:"gpu_delta,omitempty"`
	VRAMDelta   float64 `json:"vram_delta,omitempty"`
}

ResourceImpact represents the observed impact of a task on system resources.

func ConvertImpactBack 

func ConvertImpactBack(impact *decision.ResourceImpact) *ResourceImpact

ConvertImpactBack converts decision.ResourceImpact to learning.ResourceImpact.

type StatsObserver 

type StatsObserver func(task string, stats *TaskStats)

StatsObserver is called when task statistics are updated.

type TaskRecord 

type TaskRecord struct {
	Task       string          `json:"task"`
	Complexity int             `json:"complexity,omitempty"`
	StartedAt  time.Time       `json:"started_at"`
	Impact     *ResourceImpact `json:"impact,omitempty"`
}

TaskRecord represents a single task execution record.

type TaskStartRequest 

type TaskStartRequest struct {
	Task       string `json:"task"`
	Complexity int    `json:"complexity,omitempty"`
}

TaskStartRequest represents the request body for POST /task/notify.

type TaskStartResponse 

type TaskStartResponse struct {
	Received bool   `json:"received"`
	Task     string `json:"task"`
}

TaskStartResponse represents the response for POST /task/notify.

type TaskStats 

type TaskStats struct {
	Task         string  `json:"task"`
	Count        int64   `json:"count"`
	AvgCPUDelta  float64 `json:"avg_cpu_delta"`
	AvgMemDelta  float64 `json:"avg_mem_delta"`
	AvgGPUDelta  float64 `json:"avg_gpu_delta,omitempty"`
	AvgVRAMDelta float64 `json:"avg_vram_delta,omitempty"`
}

TaskStats holds aggregated statistics for a specific task type.

Source Files

View all Source files

Jump to

Keyboard shortcuts

? : This menu
/ : Search site
f or F : Jump to
y or Y : Canonical URL
go.dev uses cookies from Google to deliver and enhance the quality of its services and to analyze traffic. Learn more.