Change8

b8979

📦 llama-cppView on GitHub →
1 features

Summary

This release introduces a fused CUDA kernel operation for SSM_CONV + ADD(bias) + SILU. It also provides updated binary distributions for numerous operating systems and hardware architectures.

✨ New Features

  • CUDA kernel fused SSM_CONV + ADD(bias) + SILU operation for potential performance improvement.