1#![forbid(unsafe_code)]
8#![warn(missing_docs)]
9
10pub mod admission;
12pub mod artifact;
14pub mod attention;
16pub mod automatic;
18pub mod backend;
20pub mod cache;
22pub mod capability;
24pub mod checkpoint;
26pub mod consensus;
28pub mod execution;
30pub mod generation;
32pub mod input;
34pub mod inspection;
36pub mod media;
38pub mod observation;
40pub mod realtime;
42pub mod residency;
44pub mod scheduler;
46pub mod session_authority;
48pub mod speculative;
50pub mod topology;
52
53pub use admission::{
54 admit_preparation, ArchitecturePreparationCapabilities, PreparationAdmission,
55 PreparationAdmissionError, PreparationAdmissionRequest, PreparationMechanismCapabilities,
56};
57pub use artifact::{
58 gguf_u32_metadata_values, inspect_artifact, plan_model_preparation, resolve_gguf_companions,
59 validate_preparation_policy, ArtifactFormat, ArtifactInspection, GgufCompanionEncoding,
60 GgufCompanionRequirement, GgufCompanionRole, LoadingProtocol, MaterializationRoute,
61 ModelArtifact, ModelConfiguration, ModelConfigurationResolver, ModelPreparationPlan,
62 PreparationPolicy, QuantizationRequest, ResidencyRequest, ResolvedModelConfiguration,
63 ValidatedGguf, ValidatedGgufCompanion,
64};
65pub use attention::{AttentionPolicy, LayerSchedule, LayerScheduleError};
66pub use automatic::{
67 realize_execution_plan_drafting, realize_execution_plan_target, select_execution_plan_drafting,
68 select_execution_plan_target, speculative_decoding_telemetry, AllocatorTelemetry,
69 AutomaticPlanRequest, AutomaticPlanner, AutomaticPlannerPolicy, AutomaticPlanningBackend,
70 AutomaticPlanningError, BoundedResidencyRequirement, CandidateAdmission, DurationSeconds,
71 ExecutionPlanBackendFactory, ExecutionPlanReport, ExecutionPlanTarget,
72 ExecutionPlanTargetLoadError, ExecutionPlanTargetSelection, ExecutionTelemetry,
73 ExpertCacheTelemetry, ExternalDraftArtifact, HardwareBackendProfile, HardwareDeviceProfile,
74 HardwareMemorySemantics, HardwareProfile, ModelResourceProfile, ObservationKind, Observed,
75 PlanExplanation, PlanExplanationEntry, PlanExplanationLevel, PreparedExecutionPlanTarget,
76 RealizedDrafting, ResidencyTelemetry, RetainedAutomaticPlan, SelectedExecutionPlanDrafting,
77 SelectedExecutionPlanTarget, SpeculativeDecodingTelemetry, TimingTelemetry,
78 TokenizerCompatibilityError, TokenizerCompatibilityProof, TransferTelemetry,
79 AUTOMATIC_SCHEMA_VERSION,
80};
81pub use backend::{
82 load_model, prepare_inspected_model, BackendDescriptor, BackendError, BackendProvider,
83 BackendSession, BoundedCompletion, BoundedCompletionOutcome, BoundedCompletionWait,
84 BoundedCompletionWaitError, BoundedSubmissionOutcome, CollectiveGroupDescriptor,
85 CollectiveGroupId, CollectiveScope, Completion, CompletionCancellationMode,
86 ControlledTextGeneration, ControlledTextGenerationError, ControlledToken, DeviceCapabilities,
87 DeviceDescriptor, DistributedBackend, DistributedCapabilities, DistributedCommitEpoch,
88 DistributedCommitOutcome, DistributedCommitPhase, DistributedSession,
89 DistributedSessionDescriptor, InspectableBackendSession, ModelCapabilityBackend,
90 ModelLoadError, ModelLoadingBackend, ModelRuntime, MultimodalPreparationBackend,
91 MultimodalPreparationFailure, PreparedModel, SelectedModelPreparation, SessionCapabilities,
92 SessionCapabilityError, SpeculativeTokenFilterController, Submission, TextGeneration,
93 TextGenerationBackend, TextGenerationConfig, TextSamplingStrategy, TokenFilter,
94 TokenFilterController, TokenFilterError, TokenOutput, ValueDescriptor,
95};
96pub use capability::{
97 apply_admission_policy, estimate_runtime_state, Admission, AdmissionRejection,
98 AdmissionRequest, AdmissionResult, AvailableMemory, CacheStateStrategy, CapabilityError,
99 EstimationCompleteness, InputModalities, InputTokenCount, ModelCapabilities,
100 PhysicalMemorySemantics, RuntimeStateEstimate, SlidingWindowLayerCount, StateMemoryAssumptions,
101 StateMemoryLayout, StaticMemoryReport,
102};
103pub use execution::{
104 BackendId, DevicePlan, DraftPlacementPlan, DraftingPlan, ExecutionPlan, ExecutionPlanError,
105 ExpertCachePlan, ResidencyPlan, WeightTransformationPlan, DEFAULT_MAX_CACHED_SHARDS,
106 EXECUTION_PLAN_SCHEMA_VERSION,
107};
108pub use generation::{
109 resolve_generation_config, resolve_optimistic_reuse, CheckpointGenerationConfig, FinishReason,
110 GenerationCancellationToken, GenerationConfigOverrides, GenerationError, GenerationSequence,
111 OptimisticReuseDecision, ResolvedGenerationConfig, SemanticEvent,
112 SpeculativeCancellationDisposition, SpeculativeCommitPlan, SpeculativeConfig,
113 SpeculativeRequestId, SpeculativeRequestLifecycle, SpeculativeRequestStatus, SpeculativeRound,
114 SpeculativeSchedulerOptions, SpeculativeTail, TokenCommit, TokenTerminalSignals,
115};
116pub use input::{
117 InputExtent, InputMetadataKey, InputModality, InputPartDescriptor, InputPayloadKind,
118 InputTensorIdentity, PreparedInputError, PreparedInputIdentity,
119};
120pub use inspection::{
121 assemble_portable_model_inspection, finalize_realized_model_inspection,
122 media_feature_readiness, record_media_projector_inspection, record_processor_inspection,
123 reject_portable_artifact_inspection, ArtifactModality, ArtifactTensorEncoding, InspectionIssue,
124 InspectionIssueCode, InspectionReadiness, InspectionRequirement, InspectionSeverity,
125 MediaFeatureAvailability, MediaProjectorRequirement, ModelInspectionReport,
126 RealizedInspectionOutcomes,
127};
128pub use media::{
129 Audio, Media, MediaBinding, MediaRequestError, MultimodalRequest, MultimodalSegment, RgbImage,
130 TokenizedMultimodalRequest, TokenizedMultimodalSegment, Video, VideoSampling,
131};
132pub use observation::{
133 InspectedOutput, ObservationError, ObservationRequest, ObservationSelector, ObservationSet,
134 ObservationValue, TensorObservation, TensorObservationData,
135 AUDIO_PROJECTOR_OUTPUT_OBSERVATION_PATH, MODALITY_MERGE_OUTPUT_OBSERVATION_PATH,
136 MODEL_LOGITS_OBSERVATION_PATH, PROCESSOR_OUTPUT_OBSERVATION_PATH,
137 VISION_PROJECTOR_OUTPUT_OBSERVATION_PATH,
138};
139pub use realtime::{
140 RealtimeConfigError, RealtimeDecisionDiagnostics, RealtimeError, RealtimeForcedSource,
141 RealtimeFrameConvention, RealtimeFrameForcing, RealtimeFrameScheduleState, RealtimeFrameSlot,
142 RealtimeFrameTransition, RealtimeInputDescriptorError, RealtimeInputFrame, RealtimeOutputFrame,
143 RealtimeSampling, RealtimeScheduleError, RealtimeSlotCoordinate, RealtimeSlotOccupancy,
144 RealtimeSpeechConfig, RealtimeTargetDecision, RealtimeTargetSource, RealtimeTemporalSource,
145 MAX_REALTIME_FRAME_DELAY,
146};
147pub use residency::{
148 BackgroundPrefetchReport, PrefetchAdmission, PrefetchCompletion, PrefetchDemandObservation,
149 PrefetchDemandResolution, PrefetchExecutionState, PrefetchStateError, PrefetchWork,
150};
151pub use session_authority::{
152 SessionAdmission, SessionAdmissionError, SessionAuthority, SessionAuthorityError,
153 SubmissionLease,
154};
155pub use speculative::{
156 cancel_pending_verification, decide_speculative_proposal, propose_block,
157 resolve_commit_and_publish, resolve_optimistic_branch, resolve_round,
158 speculative_acceptance_probability, submit_verification_transaction,
159 CompletedSpeculativeRequest, CompletedSpeculativeSchedule, PendingSpeculativeVerification,
160 PreparedSpeculativeLane, ProposalDecision, PublishedSpeculativeResult,
161 PublishedSpeculativeVerification, ResolvedSpeculativeRound, SamplingPlacement,
162 SpeculativeAction, SpeculativeCallbackPublisher, SpeculativeCandidate, SpeculativeCapability,
163 SpeculativeCommit, SpeculativeConstraint, SpeculativeContinuation, SpeculativeDraft,
164 SpeculativeDraftBlock, SpeculativeDraftRandomPosition, SpeculativeDraftSource,
165 SpeculativeDriverError, SpeculativeExecutionTopology, SpeculativeExecutor,
166 SpeculativeGenerationBackend, SpeculativeGenerationBatchOutput,
167 SpeculativeGenerationBatchRequest, SpeculativeGenerationLane, SpeculativeGenerationOutput,
168 SpeculativeGenerationVisitor, SpeculativeLifecycleObserver, SpeculativeLifecycleStage,
169 SpeculativeOptimisticBranch, SpeculativeOutputError, SpeculativeOutputRuntime,
170 SpeculativePrefill, SpeculativeProposal, SpeculativePublicationStatus, SpeculativePublisher,
171 SpeculativeRandomness, SpeculativeRequest, SpeculativeRequestTable, SpeculativeSampling,
172 SpeculativeSchedule, SpeculativeSchedulerStats, SpeculativeSemanticConstraint,
173 SpeculativeSemanticState, SpeculativeStats, SpeculativeTelemetry,
174};
175pub use topology::{
176 balanced_contiguous_range, ParallelAxis, ParallelCoordinates, ParallelRankTopology,
177 ParallelTopology, SubgroupMembership, TopologyError, TopologyPreflightReport,
178};