Optionalrow: CacheOptionalcompose: (storage: Cache) => Cache
How the lego these rows came from composes the rows this layout hands
back to serial use (ComposedStorage.over).
Optional ReadonlycomposeHow the lego these rows came from composes the rows this layout hands
back to serial use (ComposedStorage.over).
ReadonlyfusedReadonlykReadonlystartReadonlystatestate() returned temporary views that the caller must release.
ReadonlytransitionReadonlyturboReadonlyvEarliest prefix whose retained representation can resume generation. Irreversible transitions may preserve bytes while invalidating older precision boundaries. Reuse policy must not trim below this offset.
ProtectedpackedProtectedpositionsA preparation method commits each request's own precision boundary. Splitting a forward for a sibling must not move that boundary.
Compiled-decode trace adapters expose the RoPE offset as an int32 array input here; real caches leave it unset (static int path).
ProtectedrowsProtectedadvanceThe committed read is the window read: a committed span reads this layout through the kernel set's fetch and the fused SDPA under the mask the layout builds, as a window does.
Decode: one new position per row (L = 1). Each row's query sees that
row's valid stored positions up to and including the one just appended,
limited to the cache's sliding window.
Window: L new positions per row, causal within the window, for prefill
chunks, prefill tails and speculative verify windows. Query i of a row
sees the row's earlier valid positions and window positions 0..i, limited
to the cache's sliding window. Left padding and prepared prefill padding are
never visible. Outputs at padded query positions are unspecified and the
caller ignores them. To append context without attending, make this call
and dispose the read unused.
Physical cache storage consumed by one additional token for one row. Recurrent caches return 0 because their state does not grow with the sequence. Optional for stateless/training adapters that are never admitted or persisted.
Apply the closure's outputs (takes ownership); returns arrays to evaluate with the step.
The closure inputs of slot, in order.
The fetch of the coming compiled step: the lego's while it holds the row,
else the rows' (decodePhaseOfRows), without moving the row.
ProtecteddecodeThe fetch a lone row's lego would choose, from this layout's positions.
A row handed back to serial use: its storage, composed again by the lego
the row came from (KvRowTransition.compose).
A row's storage, owned by the caller: what another row layout merges.
Can trim(n) drop the last n tokens? (Ring caches lose trimability
once wrapped.)
Maximum committed positions before this state changes precision.
ProtectedplainRows made for an empty layout carry the graph's read into their conversion.
Host-side bookkeeping for a 1-token step (growth, rotation), without the write.
Stable storage identity for compatibility guards and persistence. REQUIRED: every capability predicate (isPlainKvCache, isRotating*, packages/inference/src/state/persistence.ts codecs) is a string compare on this value, so a missing override used to fail OPEN into wrong routing (PRs #42/#43: BatchedRotatingCache shipped without one and batched joins dropped running rows). Wrappers report their own kind plus the inner kind.
Speculative verify-round support for NON-trimmable recurrent caches
(SSMCache — gated-DeltaNet conv + recurrent state). A cache exposing
these is spec-eligible without isTrimmable(): the serve loop arms a
round before the verify forward (the layer snapshots pre-round state —
free, MLX arrays are immutable — and records its position-local kernel
inputs), then resolves it: commit on full accept, or rollback(keep)
which restores the snapshot and bit-exactly REPLAYS the first keep
window tokens through the recurrence (identical arithmetic prefix ⇒
identical state). Trimmable caches leave these unset and keep trim().
ProtectedsyncDrop the last n tokens (future writes overwrite them). bypass (default
false → unchanged behavior) skips the isTrimmable guard for the
speculative-decode rollback case (trim the last n ≤ γ tips right after a
1 concat write); rotating caches then physically shrink the buffer tail.
The rows of the TurboQuant bf16-first lego (
Bf16FirstTurboQuantKVCache) for continuous batching: TurboQuant tensor access and value-domain handling over the shared row-transition lifecycle (when a row converts is described onTransitioningKvRows). Model projections and attention remain batched.fusedDecodeis resolved by whoever builds the layout.turboQuantKernelsis the read of the graph whose caches these rows hold (dense for Qwen 3.5, deferred for Gemma 4 and MiniCPM5); the rows this layout makes and the layout it packs them into carry it, and a converted row attends with its source's.The reads are the graph's call today: this layout's mask before the append (no sliding window), then the kernel set's fetch and attention. The dense set fetches every row decoded (
updateAndFetch) and runs the fused SDPA, Qwen 3.5's call; the deferred set fetches converted rows rotated (updateAndFetchDeferredV), runs the fused SDPA and applies the value transform that fetch captured, Gemma 4's and MiniCPM5's.