Expand description
§推論エンジンモジュール
ONNX Runtime を用いた低遅延フォワードパス、セッション管理、 および Execution Provider 解決機構を提供する。
Re-exports§
pub use batch::BatchScratchpad;pub use batch::DEFAULT_MAX_BATCH_CHUNK_SIZE;pub use cache::AlignedBuffer64;pub use cache::CacheCompressionMode;pub use cache::CachedStateFeatures;pub use cache::CachedTensorPayload;pub use cache::DEFAULT_DUAL_MAX_BATCH_CHUNK_SIZE;pub use cache::DualSessionEngine;pub use cache::StateCacheManager;pub use cache::is_aligned_64;pub use coarse::CandidateEmbeddingCache;pub use coarse::CoarseScorer;pub use coarse::CoarseToFineConfig;pub use coarse::DEFAULT_COARSE_THRESHOLD;pub use coarse::DEFAULT_NEGATIVE_KEYS;pub use coarse::DEFAULT_TOP_M_CANDIDATES;pub use coarse::EmbeddingCoarseScorer;pub use coarse::FilteredCandidates;pub use coarse::HierarchicalCoarseScorer;pub use coarse::LexicalCoarseScorer;pub use coarse::ModelDrivenCoarseScorer;pub use coarse::filter_top_candidates;pub use coarse::is_negative_candidate;pub use coarse::reconstruct_probabilities;pub use config::ExecutionProvider;pub use config::OptimizationLevel;pub use config::SessionConfig;pub use escalation::EscalationPromptBuilder;pub use escalation::EscalationTemplateConfig;pub use escalation::build_rich_escalation_prompt;pub use gating::apply_gating_to_answer;pub use gating::apply_gating_to_answer_with_energy;pub use gating::apply_gating_to_answer_with_logits;pub use gating::apply_gating_to_answers;pub use gating::apply_gating_to_answers_with_logits;pub use gating::calculate_choice_energy;pub use gating::resolve_gating_config;pub use provider::register_execution_providers;pub use session::InferenceEngine;