Skip to main content

Module automatic

Module automatic 

Source
Expand description

Backend-neutral automatic execution planning. Backend-neutral automatic execution planning.

Backends report hardware, artifact resources, and candidate admission. This module owns policy validation, resource budgeting, plan selection, feedback matching, and the serialized planning and telemetry documents.

Structs§

AllocatorTelemetry
Backend allocator observations for one execution.
AutomaticPlanRequest
Owned input to one automatic planning session.
AutomaticPlanner
Backend-neutral automatic planner.
AutomaticPlannerPolicy
Tunable, serializable automatic-planning policy.
BoundedResidencyRequirement
Exact bounded device-window requirement established by a backend probe.
CandidateAdmission
Backend candidate-admission result consumed by the neutral planner.
DurationSeconds
Floating-point duration wrapper with equality based on its bit pattern.
ExecutionPlanReport
Complete automatic-planning document suitable for JSON persistence.
ExecutionPlanTarget
One target-backend instance and retained selection realized from a portable execution plan.
ExecutionPlanTargetSelection
Backend-owned preparation selected before native target realization.
ExecutionTelemetry
Stable JSON telemetry for one completed model execution.
ExpertCacheTelemetry
Routed-expert cache occupancy summary.
ExternalDraftArtifact
Architecture-prepared assistant artifact and proven portable tokenizer compatibility.
HardwareBackendProfile
Availability and devices for one execution backend.
HardwareDeviceProfile
One logical device visible to an execution backend.
HardwareProfile
Hardware and memory observations used as automatic-planning inputs.
ModelResourceProfile
Architecture and header-derived planning facts used before a model is loaded.
PlanExplanation
Explanation accompanying a selected execution plan.
PlanExplanationEntry
One stable, machine-routable planner explanation entry.
ResidencyTelemetry
Logical bytes and transfers reported by bounded parameter residency.
RetainedAutomaticPlan
A portable report paired with the one artifact inspection used by every probe.
SelectedExecutionPlanDrafting
An external-drafting selection inseparably bound to its complete execution plan.
SelectedExecutionPlanTarget
An inspected artifact and authoritative backend selection ready for native realization.
SpeculativeDecodingTelemetry
Speculative-decoding observations for one request.
TimingTelemetry
Timings reported for one generation request.
TokenizerCompatibilityError
A target and external assistant do not share the same token-id vocabulary mapping.
TokenizerCompatibilityProof
Proof that a target and external assistant use the same token-id vocabulary mapping.
TransferTelemetry
One logical residency transfer counter.

Enums§

AutomaticPlanningError
Failure produced by portable planning or its selected backend adapter.
HardwareMemorySemantics
Serializable form of physical host/device memory semantics.
ObservationKind
Confidence attached to an observed or derived value.
Observed
A value which remains explicit when the runtime cannot produce it.
PlanExplanationLevel
Severity of one planner explanation entry.
RealizedDrafting
Backend-owned drafting resources realized for one complete execution plan.

Constants§

AUTOMATIC_SCHEMA_VERSION
Schema version shared by automatic-planning and telemetry documents.

Traits§

AutomaticPlanningBackend
High-level observations a backend supplies to the neutral planner.
ExecutionPlanBackendFactory
Selects and creates an executable whole-model backend from a portable execution plan.

Functions§

realize_execution_plan_drafting
Validates and realizes the drafting portion of a portable execution plan.
realize_execution_plan_target
Realizes native target resources for an already validated selection.
select_execution_plan_drafting
Selects drafting materialization before any native target realization.
select_execution_plan_target
Validates and selects the target portion before any native realization.
speculative_decoding_telemetry
Projects neutral speculative statistics into the stable telemetry document.

Type Aliases§

ExecutionPlanTargetLoadError
Failure while preparing the model retained by an execution-plan target.
PreparedExecutionPlanTarget
A realized backend paired with the model prepared from its retained selection.