Skip to content

Fix GPU Hang in MTPTokenIterator: flush Metal graph after verification pass #41

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

Merged
solderzzc merged 2 commits into from
May 12, 2026
Merged

Fix GPU Hang in MTPTokenIterator: flush Metal graph after verification pass #41

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e3fb72a
Fix GPU Hang: flush Metal graph with eval() after MTPTokenIterator ve…
May 12, 2026
881cddf
Potential fix for pull request finding
solderzzc May 12, 2026
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
11 changes: 10 additions & 1 deletion Libraries/MLXLMCommon/Evaluate.swift
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1227,9 +1227,18 @@ public struct MTPTokenIterator: TokenIteratorProtocol {

let mtpResult = model.callMTP(verifyInput.tokens[.newAxis], cache: cache, mtpCaches: mtpCaches)
guard !mtpResult.isEmpty else { return }

let mainLogits = mtpResult[0]

// Flush the Metal command buffer immediately after the verification forward pass.
// On hybrid SSM/attention models (e.g. Qwen35), the recurrent SSM layers accumulate
// un-evaluated graph nodes across rounds. Without an explicit sync here the Metal
// command buffer grows until it triggers the GPU Watchdog.
//
// Only force the main logits needed for verification/sampling so we avoid eagerly
// evaluating speculative MTP head logits that may be discarded on rejection.
eval(mainLogits)

Comment thread
solderzzc marked this conversation as resolved.
let mainTokens: MLXArray
var mainProcessedLogits = [MLXArray]()
if var verifyProcessor = processor {
Expand Down
Loading