mirrors/ollama-for-amd
1
0
Fork 0
mirror of https://github.com/likelovewant/ollama-for-amd.git synced 2025-12-21 14:26:30 +00:00
Code Issues Packages Projects Releases Wiki Activity

bring back sysfs based VRAM information for AMD (#12871)

Browse Source 
* build: optimize dockerfile context for iterating

This moves the copy of the source into the layer AFTER
doing software installs so we don't have to go through
the RPM install for cuda, etc. every time you touch a
source file.

* amd: implement linux sysfs based VRAM lookup

This adds a C++ implementation of sysfs DRM VRAM discovery
for more accurate free VRAM data on linux for AMD GPUs.
This commit is contained in:
Daniel Hiltgen
2025-11-17 15:40:58 -08:00
committed by GitHub
parent 399eacf486
commit 2f36d769aa
5 changed files with 186 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
Dockerfile
View File

@@ -39,14 +39,14 @@ ENV CC=clang CXX=clang++
FROM base-${TARGETARCH} AS base
ARG CMAKEVERSION
RUN curl -fsSL https://github.com/Kitware/CMake/releases/download/v${CMAKEVERSION}/cmake-${CMAKEVERSION}-linux-$(uname -m).tar.gz | tar xz -C /usr/local --strip-components 1
COPY CMakeLists.txt CMakePresets.json .
COPY ml/backend/ggml/ggml ml/backend/ggml/ggml
ENV LDFLAGS=-s
FROM base AS cpu
RUN dnf install -y gcc-toolset-11-gcc gcc-toolset-11-gcc-c++
ENV PATH=/opt/rh/gcc-toolset-11/root/usr/bin:$PATH
ARG PARALLEL
COPY CMakeLists.txt CMakePresets.json .
COPY ml/backend/ggml/ggml ml/backend/ggml/ggml
RUN --mount=type=cache,target=/root/.ccache \
cmake --preset 'CPU' \
&& cmake --build --parallel ${PARALLEL} --preset 'CPU' \
@@ -57,6 +57,8 @@ ARG CUDA11VERSION=11.8
RUN dnf install -y cuda-toolkit-${CUDA11VERSION//./-}
ENV PATH=/usr/local/cuda-11/bin:$PATH
ARG PARALLEL
COPY CMakeLists.txt CMakePresets.json .
COPY ml/backend/ggml/ggml ml/backend/ggml/ggml
RUN --mount=type=cache,target=/root/.ccache \
cmake --preset 'CUDA 11' \
&& cmake --build --parallel ${PARALLEL} --preset 'CUDA 11' \
@@ -67,6 +69,8 @@ ARG CUDA12VERSION=12.8
RUN dnf install -y cuda-toolkit-${CUDA12VERSION//./-}
ENV PATH=/usr/local/cuda-12/bin:$PATH
ARG PARALLEL
COPY CMakeLists.txt CMakePresets.json .
COPY ml/backend/ggml/ggml ml/backend/ggml/ggml
RUN --mount=type=cache,target=/root/.ccache \
cmake --preset 'CUDA 12' \
&& cmake --build --parallel ${PARALLEL} --preset 'CUDA 12' \
@@ -78,6 +82,8 @@ ARG CUDA13VERSION=13.0
RUN dnf install -y cuda-toolkit-${CUDA13VERSION//./-}
ENV PATH=/usr/local/cuda-13/bin:$PATH
ARG PARALLEL
COPY CMakeLists.txt CMakePresets.json .
COPY ml/backend/ggml/ggml ml/backend/ggml/ggml
RUN --mount=type=cache,target=/root/.ccache \
cmake --preset 'CUDA 13' \
&& cmake --build --parallel ${PARALLEL} --preset 'CUDA 13' \
@@ -87,6 +93,8 @@ RUN --mount=type=cache,target=/root/.ccache \
FROM base AS rocm-6
ENV PATH=/opt/rocm/hcc/bin:/opt/rocm/hip/bin:/opt/rocm/bin:/opt/rocm/hcc/bin:$PATH
ARG PARALLEL
COPY CMakeLists.txt CMakePresets.json .
COPY ml/backend/ggml/ggml ml/backend/ggml/ggml
RUN --mount=type=cache,target=/root/.ccache \
cmake --preset 'ROCm 6' \
&& cmake --build --parallel ${PARALLEL} --preset 'ROCm 6' \
@@ -118,6 +126,8 @@ RUN --mount=type=cache,target=/root/.ccache \
&& cmake --install build --component CUDA --strip --parallel ${PARALLEL}
FROM base AS vulkan
COPY CMakeLists.txt CMakePresets.json .
COPY ml/backend/ggml/ggml ml/backend/ggml/ggml
RUN --mount=type=cache,target=/root/.ccache \
cmake --preset 'Vulkan' \
&& cmake --build --parallel --preset 'Vulkan' \