vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

cuda : rename build flag to LLAMA_CUDA (#6299)

Browse source
This commit is contained in:
slaren 2024-03-26 01:16:01 +01:00 • committed by GitHub
parent b06c16ef9f
commit 280345968d
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
28 changed files with 129 additions and 115 deletions
Download patch file Download diff file Expand all files Collapse all files

4
.devops/full-cuda.Dockerfile
View file

@ -26,8 +26,8 @@ COPY . .
# Set nvcc architecture
ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
# Enable cuBLAS
ENV LLAMA_CUBLAS=1
# Enable CUDA
ENV LLAMA_CUDA=1
RUN make

22
.devops/llama-cpp-cublas.srpm.spec → .devops/llama-cpp-cuda.srpm.spec
View file

@ -12,7 +12,7 @@
# 4. OpenCL/CLBLAST support simply requires the ICD loader and basic opencl libraries.
# It is up to the user to install the correct vendor-specific support.
Name: llama.cpp-cublas
Name: llama.cpp-cuda
Version: %( date "+%%Y%%m%%d" )
Release: 1%{?dist}
Summary: CPU Inference of LLaMA model in pure C/C++ (no CUDA/OpenCL)
@ -32,16 +32,16 @@ CPU inference for Meta's Lllama2 models using default options.
%setup -n llama.cpp-master
%build
make -j LLAMA_CUBLAS=1
make -j LLAMA_CUDA=1
%install
mkdir -p %{buildroot}%{_bindir}/
cp -p main %{buildroot}%{_bindir}/llamacppcublas
cp -p server %{buildroot}%{_bindir}/llamacppcublasserver
cp -p simple %{buildroot}%{_bindir}/llamacppcublassimple
cp -p main %{buildroot}%{_bindir}/llamacppcuda
cp -p server %{buildroot}%{_bindir}/llamacppcudaserver
cp -p simple %{buildroot}%{_bindir}/llamacppcudasimple
mkdir -p %{buildroot}/usr/lib/systemd/system
%{__cat} <<EOF > %{buildroot}/usr/lib/systemd/system/llamacublas.service
%{__cat} <<EOF > %{buildroot}/usr/lib/systemd/system/llamacuda.service
[Unit]
Description=Llama.cpp server, CPU only (no GPU support in this build).
After=syslog.target network.target local-fs.target remote-fs.target nss-lookup.target
@ -49,7 +49,7 @@ After=syslog.target network.target local-fs.target remote-fs.target nss-lookup.t
[Service]
Type=simple
EnvironmentFile=/etc/sysconfig/llama
ExecStart=/usr/bin/llamacppcublasserver $LLAMA_ARGS
ExecStart=/usr/bin/llamacppcudaserver $LLAMA_ARGS
ExecReload=/bin/kill -s HUP $MAINPID
Restart=never
@ -67,10 +67,10 @@ rm -rf %{buildroot}
rm -rf %{_builddir}/*
%files
%{_bindir}/llamacppcublas
%{_bindir}/llamacppcublasserver
%{_bindir}/llamacppcublassimple
/usr/lib/systemd/system/llamacublas.service
%{_bindir}/llamacppcuda
%{_bindir}/llamacppcudaserver
%{_bindir}/llamacppcudasimple
/usr/lib/systemd/system/llamacuda.service
%config /etc/sysconfig/llama
%pre

4
.devops/main-cuda.Dockerfile
View file

@ -20,8 +20,8 @@ COPY . .
# Set nvcc architecture
ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
# Enable cuBLAS
ENV LLAMA_CUBLAS=1
# Enable CUDA
ENV LLAMA_CUDA=1
RUN make

2
.devops/nix/package.nix
View file

@ -192,7 +192,7 @@ effectiveStdenv.mkDerivation (
(cmakeBool "CMAKE_SKIP_BUILD_RPATH" true)
(cmakeBool "LLAMA_BLAS" useBlas)
(cmakeBool "LLAMA_CLBLAST" useOpenCL)
(cmakeBool "LLAMA_CUBLAS" useCuda)
(cmakeBool "LLAMA_CUDA" useCuda)
(cmakeBool "LLAMA_HIPBLAS" useRocm)
(cmakeBool "LLAMA_METAL" useMetalKit)
(cmakeBool "LLAMA_MPI" useMpi)

4
.devops/server-cuda.Dockerfile
View file

@ -20,8 +20,8 @@ COPY . .
# Set nvcc architecture
ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
# Enable cuBLAS
ENV LLAMA_CUBLAS=1
# Enable CUDA
ENV LLAMA_CUDA=1
RUN make