b9047
📦 llama-cppView on GitHub →
🐛 3 fixes🔧 1 symbols
Summary
This release includes several fixes related to device memory fitting, ensuring that unknown devices or non-GPU devices handle memory allocation correctly, specifically by preserving host fallback and keeping unknown GPU fit memory at zero.
🐛 Bug Fixes
- Prevent fitting to unknown device memory in common logic.
- Preserve host fallback when fitting devices are not GPUs.
- Keep unknown GPU fit memory at zero.