LM/llama.cpp for CUDA: Revision history

Jump to navigation Jump to search

Diff selection: Mark the radio buttons of the revisions to compare and hit enter or the button at the bottom.
Legend: (cur) = difference with latest revision, (prev) = difference with preceding revision, m = minor edit.

4 September 2026

  • curprev 04:0104:01, 4 September 2026‎ Tacoball talk contribs‎ 382 bytes −106‎ →‎Build
  • curprev 03:4203:42, 4 September 2026‎ Tacoball talk contribs‎ 488 bytes +488‎ 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..."
Retrieved from "https://wiki.fundamental-ramen.com/wiki/LM/llama.cpp_for_CUDA"