Skip to content

Repository navigation

Bind additionalStates on the pipelined constrained-generation path (#316) #338

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
stikves wants to merge 1 commit into from
Closed

Bind additionalStates on the pipelined constrained-generation path (#316) #338

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions swift/Sources/CoreAILanguageModels/InferenceEngines/CoreAIPipelinedEngine.swift
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1562,6 +1562,9 @@ private struct EngineImpl: ~Copyable {
var asyncStates = InferenceFunction.AsyncMutableViews()
asyncStates.insert(&keyState, for: keyCacheName)
asyncStates.insert(&valState, for: valueCacheName)
// Non-growing states (e.g. sliding-ring caches) are bound here too, like
// the plain-decode and prefill paths; nil for standard two-state bundles.
additionalStates?.bind(into: &asyncStates)

// Safe: constrained loop awaits each token before encoding the next step, so logits are consumed before overwrite.
let logitsBuffer = logits.metalBuffer
Expand Down