llama.cpp
a818f302 - CUDA: use MMQ instead of cuBLAS by default (#8075)
Commit
View On GitHub
Commit
350 days ago
CUDA: use MMQ instead of cuBLAS by default (#8075)
Parents
d62e4aaa
  • File
    CMakeLists.txt
  • File
    Makefile
  • File
    README.md
  • File
    ggml-cuda.cu
  • ggml-cuda
    • File
      common.cuh
    • File
      mmq.cu
    • File
      mmq.cuh
    • File
      mmvq.cuh
FAQ Terms Privacy Refunds Impressum