Expand description
Higher-level inference memoization cache keyed by (expr_fingerprint, input_hash).
This module provides a memoization layer distinct from the lower-level TensorCache
in cache.rs. Entries are keyed by a MemoKey that combines an expression fingerprint
(derived from a TLExpr) with an optional hash over input values.
Structs§
- Memo
Cache - A memoization cache for inference results keyed by
MemoKey. - Memo
Cache Builder - Builder for
MemoCache, following the COOLJAPAN builder pattern. - Memo
Config - Configuration for a
MemoCache. - MemoKey
- Key for memoized results.
- Memo
Stats - Runtime statistics for a
MemoCache.
Enums§
- Memo
Eviction Policy - Eviction policy for the memo cache.
- Memo
Lookup Result - The outcome of a
MemoCache::getcall.
Type Aliases§
- Expr
Memo Cache - Type alias for the primary use-case: caching
ArrayD<f64>inference results.