LM/llama.cpp for CUDA: Difference between revisions

From Fundamental Ramen
< LM
Jump to navigation Jump to search
(Created page with "== Build == <syntaxhighlight lang="bash"> cmake -B build/cuda -G Ninja \ -DGGML_CUDA=ON \ -DCMAKE_BUILD_TYPE=Release cmake -B build/cuda -G Ninja \ -DGGML_CUDA=ON \ -DGGML_NATIVE=ON \ -DGGML_LTO=ON \ -DCMAKE_BUILD_TYPE=Release \ -DCMAKE_INSTALL_RPATH="\$ORIGIN/../lib" \ -DCMAKE_CXX_FLAGS="-O3 -fp-model=fast=2 -qopenmp" \ -DCMAKE_C_FLAGS="-O3 -fp-model=fast=2 -qopenmp" cmake --build build/cuda cmake --install build/cuda --prefix ~/llama.cpp/cuda </syn...")
 
 
Line 11: Line 11:
   -DGGML_LTO=ON \
   -DGGML_LTO=ON \
   -DCMAKE_BUILD_TYPE=Release \
   -DCMAKE_BUILD_TYPE=Release \
   -DCMAKE_INSTALL_RPATH="\$ORIGIN/../lib" \
   -DCMAKE_INSTALL_RPATH="\$ORIGIN/../lib"
  -DCMAKE_CXX_FLAGS="-O3 -fp-model=fast=2 -qopenmp" \
  -DCMAKE_C_FLAGS="-O3 -fp-model=fast=2 -qopenmp"


cmake --build build/cuda
cmake --build build/cuda
cmake --install build/cuda --prefix ~/llama.cpp/cuda
cmake --install build/cuda --prefix ~/llama.cpp/cuda
</syntaxhighlight>
</syntaxhighlight>

Latest revision as of 04:01, 4 September 2026

Build

cmake -B build/cuda -G Ninja \
  -DGGML_CUDA=ON \
  -DCMAKE_BUILD_TYPE=Release

cmake -B build/cuda -G Ninja \
  -DGGML_CUDA=ON \
  -DGGML_NATIVE=ON \
  -DGGML_LTO=ON \
  -DCMAKE_BUILD_TYPE=Release \
  -DCMAKE_INSTALL_RPATH="\$ORIGIN/../lib"

cmake --build build/cuda
cmake --install build/cuda --prefix ~/llama.cpp/cuda
Retrieved from "https://wiki.fundamental-ramen.com/index.php?title=LM/llama.cpp_for_CUDA&oldid=3246"