Change8

b7590

Breaking Changes
📦 llama-cppView on GitHub →
1 breaking🔧 2 symbols

Summary

This release removes BF16 x F16 kernels from the Metal backend and provides updated binaries across multiple platforms and hardware accelerators.

⚠️ Breaking Changes

  • Metal: Removed BF16 x F16 kernels. This may affect performance or compatibility for specific mixed-precision operations on Apple Silicon.

Migration Steps

  1. If using Metal for inference, ensure your model weights and compute precision are compatible with the remaining F16 or BF16 kernel paths.

🔧 Affected Symbols

ggml-metal.mmetal