commit: ee54de8ccc9acd462f837185e28ec624553d2a95 Author: Paul Zander <negril.nx+gentoo <AT> gmail <DOT> com> AuthorDate: Sat Feb 8 19:30:07 2025 +0000 Commit: Florian Schmaus <flow <AT> gentoo <DOT> org> CommitDate: Sat Feb 8 21:42:29 2025 +0000 URL: https://gitweb.gentoo.org/repo/proj/guru.git/commit/?id=ee54de8c
app-misc/ollama: update 9999 Closes: https://bugs.gentoo.org/948424 Closes: https://bugs.gentoo.org/935842 Closes: https://bugs.gentoo.org/920301 Signed-off-by: Paul Zander <negril.nx+gentoo <AT> gmail.com> .../ollama/files/ollama-9999-include-cstdint.patch | 21 ++ app-misc/ollama/metadata.xml | 1 + app-misc/ollama/ollama-9999.ebuild | 293 ++++++++++++++++----- 3 files changed, 254 insertions(+), 61 deletions(-) diff --git a/app-misc/ollama/files/ollama-9999-include-cstdint.patch b/app-misc/ollama/files/ollama-9999-include-cstdint.patch new file mode 100644 index 000000000..906d194bf --- /dev/null +++ b/app-misc/ollama/files/ollama-9999-include-cstdint.patch @@ -0,0 +1,21 @@ +From 7aa9bca5dfd47bd2c88726d53c07107f9a5daccf Mon Sep 17 00:00:00 2001 +From: Paul Zander <[email protected]> +Date: Fri, 31 Jan 2025 22:54:29 +0100 +Subject: [PATCH] include cstdint + +Signed-off-by: Paul Zander <[email protected]> + +diff --git a/llama/llama.cpp/src/llama-mmap.h b/llama/llama.cpp/src/llama-mmap.h +index 6bcddee8..8a63b8fd 100644 +--- a/llama/llama.cpp/src/llama-mmap.h ++++ b/llama/llama.cpp/src/llama-mmap.h +@@ -1,5 +1,6 @@ + #pragma once + ++#include <cstdint> + #include <memory> + #include <vector> + +-- +2.48.1 + diff --git a/app-misc/ollama/metadata.xml b/app-misc/ollama/metadata.xml index cb4ddec8a..7582e2eee 100644 --- a/app-misc/ollama/metadata.xml +++ b/app-misc/ollama/metadata.xml @@ -8,6 +8,7 @@ </maintainer> <use> <flag name="cuda">Enable NVIDIA CUDA support</flag> + <flag name="mkl">Use <pkg>sci-libs/mkl</pkg> for blas, lapack and sparse blas routines</flag> <flag name="rocm">Enable ROCm gpu computing support</flag> </use> <upstream> diff --git a/app-misc/ollama/ollama-9999.ebuild b/app-misc/ollama/ollama-9999.ebuild index f0d2f0be2..79153a59f 100644 --- a/app-misc/ollama/ollama-9999.ebuild +++ b/app-misc/ollama/ollama-9999.ebuild @@ -4,61 +4,153 @@ EAPI=8 ROCM_VERSION=6.1 -inherit git-r3 go-module rocm +inherit cuda rocm +inherit cmake +inherit go-module toolchain-funcs DESCRIPTION="Get up and running with Llama 3, Mistral, Gemma, and other language models." HOMEPAGE="https://ollama.com"; -EGIT_REPO_URI="https://github.com/ollama/ollama.git"; + +if [[ ${PV} == *9999* ]]; then + inherit git-r3 + EGIT_REPO_URI="https://github.com/ollama/ollama.git"; +else + KEYWORDS="~amd64" +fi + LICENSE="MIT" SLOT="0" -IUSE="cuda video_cards_amdgpu -cpu_flags_x86_avx cpu_flags_x86_avx2 -cpu_flags_x86_avx512f cpu_flags_x86_avx512vbmi cpu_flags_x86_avx512_vnni cpu_flags_x86_avx512_bf16 +X86_CPU_FLAGS=( + avx + f16c + avx2 + fma3 + avx512f + avx512vbmi + avx512_vnni + avx512_bf16 + avx_vnni + amx_tile + amx_int8 +) +CPU_FLAGS=( "${X86_CPU_FLAGS[@]/#/cpu_flags_x86_}" ) +IUSE="${CPU_FLAGS[*]} cuda blas mkl rocm" +# IUSE+=" opencl vulkan" + +COMMON_DEPEND=" + cuda? ( + dev-util/nvidia-cuda-toolkit:= + ) + blas? ( + !mkl? ( + virtual/blas + ) + mkl? ( + sci-libs/mkl + ) + ) + rocm? ( + >=sci-libs/hipBLAS-${ROCM_VERSION}:=[${ROCM_USEDEP}] + ) " -REQUIRED_USE=" - cpu_flags_x86_avx2? ( cpu_flags_x86_avx ) - cpu_flags_x86_avx512f? ( cpu_flags_x86_avx2 ) - cpu_flags_x86_avx512vbmi? ( cpu_flags_x86_avx512f ) - cpu_flags_x86_avx512_vnni? ( cpu_flags_x86_avx512f ) - cpu_flags_x86_avx512_bf16? ( cpu_flags_x86_avx512f ) +DEPEND=" + ${COMMON_DEPEND} + >=dev-lang/go-1.23.4 " RDEPEND=" - acct-group/ollama - acct-user/ollama -" -IDEPEND="${RDEPEND}" -BDEPEND=" - >=dev-lang/go-1.23.4 - >=dev-build/cmake-3.24 - >=sys-devel/gcc-11.4.0 - cuda? ( dev-util/nvidia-cuda-toolkit ) - video_cards_amdgpu? ( - =sci-libs/hipBLAS-${ROCM_VERSION}* - ) + ${COMMON_DEPEND} + acct-group/${PN} + acct-user/${PN} " -pkg_pretend() { - if use video_cards_amdgpu || use cuda; then - ewarn "WARNING: AMD & Nvidia support in this ebuild are experimental" - einfo "If you run into issues, especially compiling dev-libs/rocm-opencl-runtime" - einfo "you may try the docker image here https://github.com/ROCm/ROCm-docker"; - einfo "and follow instructions here" - einfo "https://rocm.docs.amd.com/projects/install-on-linux/en/latest/how-to/docker.html"; - fi -} +PATCHES=( + "${FILESDIR}/${PN}-9999-include-cstdint.patch" +) src_unpack() { - git-r3_src_unpack - go-module_live_vendor + if [[ "${PV}" == *9999* ]]; then + git-r3_src_unpack + go-module_live_vendor + else + go-module_src_unpack + fi } src_prepare() { - default + cmake_src_prepare - if use video_cards_amdgpu; then + sed -e "/set(GGML_CCACHE/s/ON/OFF/g" -i CMakeLists.txt || die + + if use amd64; then + if ! use cpu_flags_x86_avx; then + sed -e "/ggml_add_cpu_backend_variant(sandybridge/s/^/# /g" -i ml/backend/ggml/ggml/src/CMakeLists.txt || die + # AVX) + fi + if + ! use cpu_flags_x86_avx || + ! use cpu_flags_x86_f16c || + ! use cpu_flags_x86_avx2 || + ! use cpu_flags_x86_fma3; then + sed -e "/ggml_add_cpu_backend_variant(haswell/s/^/# /g" -i ml/backend/ggml/ggml/src/CMakeLists.txt || die + # AVX F16C AVX2 FMA) + fi + if + ! use cpu_flags_x86_avx || + ! use cpu_flags_x86_f16c || + ! use cpu_flags_x86_avx2 || + ! use cpu_flags_x86_fma3 || + ! use cpu_flags_x86_avx512f; then + sed -e "/ggml_add_cpu_backend_variant(skylakex/s/^/# /g" -i ml/backend/ggml/ggml/src/CMakeLists.txt || die + # AVX F16C AVX2 FMA AVX512) + fi + if + ! use cpu_flags_x86_avx || + ! use cpu_flags_x86_f16c || + ! use cpu_flags_x86_avx2 || + ! use cpu_flags_x86_fma3 || + ! use cpu_flags_x86_avx512f || + ! use cpu_flags_x86_avx512vbmi || + ! use cpu_flags_x86_avx512_vnni; then + sed -e "/ggml_add_cpu_backend_variant(icelake/s/^/# /g" -i ml/backend/ggml/ggml/src/CMakeLists.txt || die + # AVX F16C AVX2 FMA AVX512 AVX512_VBMI AVX512_VNNI) + fi + if + ! use cpu_flags_x86_avx || + ! use cpu_flags_x86_f16c || + ! use cpu_flags_x86_avx2 || + ! use cpu_flags_x86_fma3 || + ! use cpu_flags_x86_avx_vnni; then + sed -e "/ggml_add_cpu_backend_variant(alderlake/s/^/# /g" -i ml/backend/ggml/ggml/src/CMakeLists.txt || die + # AVX F16C AVX2 FMA AVX_VNNI) + fi + + if + ! use cpu_flags_x86_avx || + ! use cpu_flags_x86_f16c || + ! use cpu_flags_x86_avx2 || + ! use cpu_flags_x86_fma3 || + ! use cpu_flags_x86_avx512f || + ! use cpu_flags_x86_avx512vbmi || + ! use cpu_flags_x86_avx512_vnni || + ! use cpu_flags_x86_avx512_bf16 || + ! use cpu_flags_x86_amx_tile || + ! use cpu_flags_x86_amx_int8 ; then + sed -e "/ggml_add_cpu_backend_variant(sapphirerapids/s/^/# /g" -i ml/backend/ggml/ggml/src/CMakeLists.txt || die + #AVX F16C AVX2 FMA AVX512 AVX512_VBMI AVX512_VNNI AVX512_BF16 AMX_TILE AMX_INT8) + fi + : # ml/backend/ggml/ggml/src/CMakeLists.txt + fi + + # default + # return + if use cuda; then + cuda_src_prepare + fi + + if use rocm; then # --hip-version gets appended to the compile flags which isn't a known flag. # This causes rocm builds to fail because -Wunused-command-line-argument is turned on. # Use nuclear option to fix this. @@ -67,45 +159,124 @@ src_prepare() { fi } -src_compile() { - CUSTOM_CPU_FLAGS="" - use cpu_flags_x86_avx && CUSTOM_CPU_FLAGS+="avx" - use cpu_flags_x86_avx2 && CUSTOM_CPU_FLAGS+=",avx2" - use cpu_flags_x86_avx512f && CUSTOM_CPU_FLAGS+=",avx512" - use cpu_flags_x86_avx512vbmi && CUSTOM_CPU_FLAGS+=",avx512vbmi" - use cpu_flags_x86_avx512_vnni && CUSTOM_CPU_FLAGS+=",avx512vnni" - use cpu_flags_x86_avx512_bf16 && CUSTOM_CPU_FLAGS+=",avx512bf16" - - # Build basic ollama executable with cpu features built in - export CUSTOM_CPU_FLAGS - - if use video_cards_amdgpu; then - export HIP_ARCHS=$(get_amdgpu_flags) - export HIP_PATH="/usr" +src_configure() { + local mycmakeargs=( + -DGGML_CCACHE="no" + + -DGGML_BLAS="$(usex blas)" + # -DGGML_CUDA="$(usex cuda)" + # -DGGML_HIP="$(usex rocm)" + + # -DGGML_METAL="yes" # apple + # missing from ml/backend/ggml/ggml/src/ + # -DGGML_CANN="yes" + # -DGGML_MUSA="yes" + # -DGGML_RPC="yes" + # -DGGML_SYCL="yes" + # -DGGML_KOMPUTE="$(usex kompute)" + # -DGGML_OPENCL="$(usex opencl)" + # -DGGML_VULKAN="$(usex vulkan)" + ) + + if use blas; then + if use mkl; then + mycmakeargs+=( + -DGGML_BLAS_VENDOR="Intel" + ) + else + mycmakeargs+=( + -DGGML_BLAS_VENDOR="Generic" + ) + fi + fi + if use cuda; then + local -x CUDAHOSTCXX CUDAHOSTLD + CUDAHOSTCXX="$(cuda_gccdir)" + CUDAHOSTLD="$(tc-getCXX)" + + cuda_add_sandbox -w else - export OLLAMA_SKIP_ROCM_GENERATE=1 + mycmakeargs+=( + -DCMAKE_CUDA_COMPILER="NOTFOUND" + ) fi - if ! use cuda; then - export OLLAMA_SKIP_CUDA_GENERATE=1 + if use rocm; then + mycmakeargs+=( + -DCMAKE_HIP_PLATFORM="amd" + ) + local -x HIP_ARCHS=$(get_amdgpu_flags) + local -x HIP_PATH="/usr" + + check_amdgpu + else + mycmakeargs+=( + -DCMAKE_HIP_COMPILER="NOTFOUND" + ) fi - emake dist + + cmake_src_configure + + # if ! use cuda && ! use rocm; then + # # to configure and build only CPU variants + # set -- cmake --preset Default "${mycmakeargs[@]}" + # fi + + # if use cuda; then + # # to configure and build only CUDA + # set -- cmake --preset CUDA "${mycmakeargs[@]}" + # fi + + # if use rocm; then + # # to configure and build only ROCm + # set -- cmake --preset ROCm "${mycmakeargs[@]}" + # fi + + # echo "$@" >&2 + # "$@" || die -n "${*} failed" +} + +src_compile() { + ego build + + cmake_src_compile + + # if ! use cuda && ! use rocm; then + # # to configure and build only CPU variants + # set -- cmake --build --preset Default -j16 + # fi + + # if use cuda; then + # # to configure and build only CUDA + # set -- cmake --build --preset CUDA -j16 + # fi + + # if use rocm; then + # # to configure and build only ROCm + # set -- cmake --build --preset ROCm -j16 + # fi + + # echo "$@" >&2 + # "$@" || die -n "${*} failed" } src_install() { - dobin dist/linux-${ARCH}/bin/ollama + dobin ollama + + cmake_src_install - if [[ -d "dist/linux-${ARCH}/lib/ollama" ]] ; then - insinto /usr/lib - doins -r dist/linux-${ARCH}/lib/ollama + if use cuda; then + # remove the copied cuda files... + rm "${ED}/usr/lib/ollama"/cuda_*/libcu*.so* || die fi - doinitd "${FILESDIR}"/ollama + doinitd "${FILESDIR}"/ollama.init } pkg_preinst() { keepdir /var/log/ollama - fowners ollama:ollama /var/log/ollama + # fowners ollama:ollama /var/log/ollama + fperms 777 /var/log/ollama } pkg_postinst() {
