go-mlx/internal
Snider 51ac442a09
Some checks failed
Security Scan / security (push) Successful in 15s
Test / Vet & Build (push) Failing after 32s
fix: add deterministic GPU memory cleanup across inference paths
- defer freeCaches() in Generate and InspectAttention
- Free orphaned arrays during KVCache growth and slice updates
- Free per-token scalar intermediates in samplers and ops
- Free intermediate arrays in applyRepeatPenalty

Found by 3-way review: Claude explorer, Codex (gpt-5.3), Gemini Ultra.
Gemini implemented the fixes.

Co-Authored-By: Gemini <noreply@google.com>
Co-Authored-By: Virgil <virgil@lethean.io>
2026-02-23 05:08:02 +00:00
..
metal fix: add deterministic GPU memory cleanup across inference paths 2026-02-23 05:08:02 +00:00