b9878
📦 llama-cppView on GitHub →
🐛 1 fixes🔧 1 symbols
Summary
This release fixes stale tensor-split parameters for draft models, specifically targeting GQA attention metadata, and includes several internal code reverts.
Migration Steps
- Revert commit b90c6c6300091fe09e2350a3d4edcfcf15db8d2e if it was previously applied, as it was reverted in this release.
🐛 Bug Fixes
- Fixed stale tensor-split parameters for draft models, specifically addressing tensor split metadata for GQA attention.