upstream-archive/ollama
1
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-23 09:10:53 -05:00
Code Issues Packages Projects Releases Wiki Activity

runner: Remove CGO engines, use llama-server exclusively for GGML models (#16031)

Browse Source 
* broad lint fixes to sidestep CI scope glitch

* runner: Remove CGO engines, use llama-server exclusively for GGML models

Remove the vendored GGML and llama.cpp backend, CGO runner, Go model
implementations, and sample.  llama-server (built from upstream llama.cpp via
FetchContent) is now the sole inference engine for GGUF-based models.
(Safetensor based models continue to run on the new MLX engine.)  This allows
us to more rapidly pick up new capabilities and fixes from llama.cpp as they
come out.

On windows this now requires recent AMD driver versions to support ROCm v7 as
llama.cpp currently does not support building against v6.

* llama/compat: load Ollama-format GGUFs in llama-server

Squashed from upstream/jmorganca/llama-compat on 2026-04-29.
Source tip: 0c33775d37.

Original source commits:
- 25223160d llama/compat: add in-memory shim so llama-server can load Ollama-format GGUFs
- 7449b539a llm,server: route Ollama-format gemma3 blobs through llama/compat
- 436f2e2b1 llama/compat: make patch-apply idempotent
- 8c2c9d4c8 llama/compat: extend gemma3 handler to cover 1B and 270M blobs
- 021389f7b llama/compat: shrink clip.cpp injection from 18 lines to 1
- 61b367ec2 llama/compat: shrink patch to pure call-site hooks (34 -> 20 lines)
- 36049361c llama/compat: simplify shim (gemma3-tested)
- 8fa664865 llama/compat: add qwen35moe text handler
- db0c74530 llama/compat: add qwen35moe vision (clip) support
- 2a388da77 llama/compat: split shared infra into a util TU
- 9a69a17dc llama/compat: document non-public API dependencies
- d0f38a915 llama/compat: add gpt-oss and lfm2 handlers
- 086071822 llama/compat: add mistral3 text handler (vision TODO)
- 63bde9ff7 llama/compat: add mistral3 vision (clip) support
- 3a57b89d5 llama/compat: apply LLaMA RoPE permute to mistral3 vision Q/K
- 99cb87439 llama/compat: add qwen35, gemma4, deepseek-ocr handlers
- 2c7850dba llama/compat: add nemotron_h_moe handler (latent FFN + MTP skip)
- 9e3b54225 llama/compat: add llama4 text + clip handlers
- 034fee349 llama/compat: add gemma4 clip handler (gemma4v projector)
- 9945c5a93 server: remove dhiltgen/* compat redirect table
- 5d4539101 llama/compat: rewrite gemma4 tokenizer model to BPE
- 7e0765327 llama/compat: add glm-ocr text handler + text-loader load-op hook
- f1bd1a25a llama/compat: add glm-ocr clip handler (glm4v projector)
- 4b5cf3420 llama/compat: collapse text-loader hook back to one new patch line
- eb4ecf4fc llama/compat: extend gemma4 clip handler to gemma4a (audio)
- a23a5e76f llama/compat: fix gemma4a per-block norm tensor mapping
- cd2dcaff4 llama/compat: add embeddinggemma handler
- 1ce8a6b26 llama/compat: add qwen3-vl + qwen2.5-vl handlers
- fd98ffa1e llama/compat: add gemma3n + glm4moelite handlers
- cc7bdf0bc llama/compat: handle null buft in maybe_load_tensor
- 0c33775d3 llama/compat: disable mmap when load_op transforms text-side tensors

* refine implementation

* ci: fix windows MLX build

* ci: fix windows llama-server build

* ci: fix windows rocm build

* ci: windows mlx tuning

Shorten long-tail on build, and get OllamaSetup.exe back under 2g limit

* ci: fix windows dependencies

* win: fix dependency gathering

* disable openmp

* win: arm64 cross-compile build

also DRY out CI steps

* scheduler improvements

* ci: improvements from #15982

* win: favor ninja for faster developer builds

* win: fix build

* win: fix arm64 cross-compile

* win: avoid spaces in compiler path

* misc discovery fixes, and bos handling

* lint fixes

* win: fix arm cross-compile build/CI bugs

* llama.cpp update

* win: handle multiple CRT dirs

* vulkan: add windows iGPU detection

* fix creation bugs for patched models, other refactoring work

* tune batch size for better performance

* ci and lint fixes

* fix repeat_last_n bug

* build: revamp build for better developer UX

* amd, sampler, qwen3next fixes

* version bump

* fix mlx build

* revamp GPU discovery

Scanning the output of llama-server is turning out to be too error prone across
llama.cpp updates, so this switches to a thin dynamic library load against the
bundled GGML libraries so more details can be gathered from the API.

* version bump

* missing file

* ci: fix cache miss on rocm build

* refine vulkan dep handling

* fix ps reporting bug on full GPU load

* improve cmake wiring for customized local builds

* version bump

* docker build arg cleanup

* improve windows exit error logs

* fix community gemma4 support and ci flakes

* fix mlx unit test

* tighten up ps logic to avoid double counting fit log lines

* version bump

* fix ps view for full gpu layer offload

* add MTP wiring for llama-server and create with GGUFs

* pick best template by capabilities

* version bump

* ci: harden apt repos

* remove unused cpu core discovery

* adjust batch default logic to reduce OOMs

* support larger tool calls

* fix audio support, template show

* qwen35 mtp patch support

* flesh out dtypes

* rocm deps

* version bump

* lint fix

* block broken gfx1150 on windows

* fix qwen3.5 moe mtp tensors in patch

* mmproj oom fallback and vulkan on by default

* qwen MTP compat fix

* version bump

* ci: fix WoA cross-compile

* ci: workaround ui tool in cross-compile

* version bump

* win: enable OpenMP for CPU builds

* build: improve developer UX

* ci: windows path workaround for CPU build

* win: fix WoA dependencies

* win: fix large offset reads for mmproj patched loads

* version bump

* fix vulkan dup detection

* add OLLAMA_IGPU_ENABLE and largely disable iGPUs by default

* opt-in MTP, win large offset, integraton fixes

* fix unit test scheduler interaction hang

* fix multi-gpu filtering

* version bump

* review comments

* fix thinking level

* fix linux rocm ordering and granite 3.3 template

* version bump

* ci fix - non-shallow MLX checkout

* bypass linux sysfs unit test on windows

---------

Co-authored-by: jmorganca <jmorganca@gmail.com>
This commit is contained in:
Daniel Hiltgen
2026-05-29 13:35:47 -07:00
committed by GitHub
parent f63eea3d27
commit 9db4bdbad6
1100 changed files with 28510 additions and 430069 deletions
Download Patch File Download Diff File Expand all files Collapse all files

236
.github/workflows/release.yaml vendored
View File

@@ -16,7 +16,7 @@ jobs:
outputs:
GOFLAGS: ${{ steps.goflags.outputs.GOFLAGS }}
VERSION: ${{ steps.goflags.outputs.VERSION }}
vendorsha: ${{ steps.changes.outputs.vendorsha }}
vendorsha: ${{ steps.goflags.outputs.vendorsha }}
steps:
- uses: actions/checkout@v4
- name: Set environment
@@ -24,7 +24,7 @@ jobs:
run: |
echo GOFLAGS="'-ldflags=-w -s \"-X=github.com/ollama/ollama/version.Version=${GITHUB_REF_NAME#v}\" \"-X=github.com/ollama/ollama/server.mode=release\"'" | tee -a $GITHUB_OUTPUT
echo VERSION="${GITHUB_REF_NAME#v}" | tee -a $GITHUB_OUTPUT
echo vendorsha=$(make -f Makefile.sync print-base) | tee -a $GITHUB_OUTPUT
echo vendorsha=$(cat LLAMA_CPP_VERSION)-$(cat MLX_VERSION)-$(cat MLX_C_VERSION) | tee -a $GITHUB_OUTPUT
darwin-build:
runs-on: macos-26-xlarge
@@ -57,7 +57,9 @@ jobs:
go-version-file: go.mod
cache-dependency-path: |
go.sum
Makefile.sync
LLAMA_CPP_VERSION
MLX_VERSION
MLX_C_VERSION
- run: |
./scripts/build_darwin.sh
- name: Log build results
@@ -73,15 +75,18 @@ jobs:
dist/*.dmg
windows-depends:
needs: setup-environment
strategy:
matrix:
os: [windows]
arch: [amd64]
preset: ['CPU']
build-steps: ['cpu cpuArm64']
include:
- os: windows
arch: amd64
preset: 'CUDA 12'
build-steps: cuda12
install: https://developer.download.nvidia.com/compute/cuda/12.8.0/local_installers/cuda_12.8.0_571.96_windows.exe
cuda-components:
- '"cudart"'
@@ -89,10 +94,10 @@ jobs:
- '"cublas"'
- '"cublas_dev"'
cuda-version: '12.8'
flags: ''
- os: windows
arch: amd64
preset: 'CUDA 13'
build-steps: cuda13
install: https://developer.download.nvidia.com/compute/cuda/13.0.0/local_installers/cuda_13.0.0_windows.exe
cuda-components:
- '"cudart"'
@@ -103,23 +108,23 @@ jobs:
- '"nvvm"'
- '"nvptxcompiler"'
cuda-version: '13.0'
flags: ''
- os: windows
arch: amd64
preset: 'ROCm 6'
install: https://download.amd.com/developer/eula/rocm-hub/AMD-Software-PRO-Edition-24.Q4-WinSvr2022-For-HIP.exe
rocm-version: '6.2'
flags: '-DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DCMAKE_C_FLAGS="-parallel-jobs=4 -Wno-ignored-attributes -Wno-deprecated-pragma" -DCMAKE_CXX_FLAGS="-parallel-jobs=4 -Wno-ignored-attributes -Wno-deprecated-pragma"'
runner_dir: 'rocm'
preset: 'ROCm 7'
build-steps: rocm7
install: https://download.amd.com/developer/eula/rocm-hub/AMD-Software-PRO-Edition-26.Q1-Win11-For-HIP.exe
rocm-version: '7.1'
- os: windows
arch: amd64
preset: Vulkan
build-steps: vulkan
install: https://sdk.lunarg.com/sdk/download/1.4.321.1/windows/vulkansdk-windows-X64-1.4.321.1.exe
flags: ''
runner_dir: 'vulkan'
- os: windows
arch: amd64
preset: 'MLX CUDA 13'
build-steps: mlxCuda13
build-parallel: '16'
cmake-cuda-flags: '-t 6'
install: https://developer.download.nvidia.com/compute/cuda/13.0.0/local_installers/cuda_13.0.0_windows.exe
cudnn-install: https://developer.download.nvidia.com/compute/cudnn/redist/cudnn/windows-x86_64/cudnn-windows-x86_64-9.18.1.3_cuda13-archive.zip
cuda-components:
@@ -135,13 +140,12 @@ jobs:
- '"nvvm"'
- '"nvptxcompiler"'
cuda-version: '13.0'
flags: ''
runs-on: ${{ matrix.arch == 'arm64' && format('{0}-{1}', matrix.os, matrix.arch) || matrix.os }}
environment: release
env:
GOFLAGS: ${{ needs.setup-environment.outputs.GOFLAGS }}
VERSION: ${{ needs.setup-environment.outputs.VERSION }}
steps:
# Increase pagefile to handle momentary spikes in RAM from NVCC compiles
- if: startsWith(matrix.preset, 'MLX ')
name: Increase pagefile to 200 GB
uses: al-cheb/configure-pagefile-action@v1.5
@@ -155,6 +159,15 @@ jobs:
if (Get-Command ccache -ErrorAction SilentlyContinue) {
ccache -o cache_dir=${{ github.workspace }}\.ccache
}
- if: matrix.preset == 'CPU'
name: Install Windows ARM64 cross compiler
run: |
Invoke-WebRequest -Uri "https://github.com/mstorsjo/llvm-mingw/releases/download/20240619/llvm-mingw-20240619-ucrt-x86_64.zip" -OutFile "${{ runner.temp }}\llvm-mingw-ucrt.zip"
Expand-Archive -Path ${{ runner.temp }}\llvm-mingw-ucrt.zip -DestinationPath "C:\Program Files\"
$installPath=(Resolve-Path -Path "C:\Program Files\llvm-mingw-*-ucrt-x86_64").path
if (!(Test-Path "$installPath\bin\aarch64-w64-mingw32-gcc.exe")) {
throw "llvm-mingw x86_64 package is missing the aarch64 cross compiler"
}
- if: startsWith(matrix.preset, 'CUDA ') || startsWith(matrix.preset, 'ROCm ') || startsWith(matrix.preset, 'Vulkan') || startsWith(matrix.preset, 'MLX ')
id: cache-install
uses: actions/cache/restore@v4
@@ -203,12 +216,12 @@ jobs:
}
$vulkanPath = (Resolve-Path "C:\VulkanSDK\*").path
$vulkanRuntime = Join-Path $vulkanPath "Helpers\VulkanRT.exe"
if (Test-Path $vulkanRuntime) {
Start-Process -FilePath $vulkanRuntime -ArgumentList "/s" -NoNewWindow -Wait
}
echo "$vulkanPath\bin" | Out-File -FilePath $env:GITHUB_PATH -Encoding utf8 -Append
echo "VULKAN_SDK=$vulkanPath" >> $env:GITHUB_ENV
- if: matrix.preset == 'CPU'
run: |
echo "CC=clang.exe" | Out-File -FilePath $env:GITHUB_ENV -Append
echo "CXX=clang++.exe" | Out-File -FilePath $env:GITHUB_ENV -Append
- if: startsWith(matrix.preset, 'MLX ')
name: Install cuDNN for MLX
run: |
@@ -240,73 +253,63 @@ jobs:
with:
path: ${{ github.workspace }}\.ccache
key: ccache-${{ matrix.os }}-${{ matrix.arch }}-${{ matrix.preset }}-${{ needs.setup-environment.outputs.vendorsha }}
- name: Build target "${{ matrix.preset }}"
- name: Build Windows dependencies
run: |
Import-Module 'C:\Program Files\Microsoft Visual Studio\2022\Enterprise\Common7\Tools\Microsoft.VisualStudio.DevShell.dll'
Enter-VsDevShell -VsInstallPath 'C:\Program Files\Microsoft Visual Studio\2022\Enterprise' -SkipAutomaticLocation -DevCmdArguments '-arch=x64 -no_logo'
cmake --preset "${{ matrix.preset }}" ${{ matrix.flags }} --install-prefix "$((pwd).Path)\dist\${{ matrix.os }}-${{ matrix.arch }}"
cmake --build --preset "${{ matrix.preset }}" -- -l $([Environment]::ProcessorCount)
cmake --install build --component "${{ startsWith(matrix.preset, 'MLX ') && 'MLX' || startsWith(matrix.preset, 'CUDA ') && 'CUDA' || startsWith(matrix.preset, 'ROCm ') && 'HIP' || startsWith(matrix.preset, 'Vulkan') && 'Vulkan' || 'CPU' }}" --strip
if ('${{ matrix.preset }}'.StartsWith('MLX ')) { cmake --install build --component MLX_VENDOR }
Remove-Item -Path dist\lib\ollama\rocm\rocblas\library\*gfx906* -ErrorAction SilentlyContinue
$steps = "${{ matrix.build-steps }}".Split(' ', [System.StringSplitOptions]::RemoveEmptyEntries)
./scripts/build_windows.ps1 @steps
env:
CMAKE_GENERATOR: Ninja
OLLAMA_BUILD_PARALLEL: ${{ matrix.build-parallel || '' }}
OLLAMA_CMAKE_CUDA_FLAGS: ${{ matrix.cmake-cuda-flags || '' }}
- name: Log build results
run: |
gci -path .\dist -Recurse -File | ForEach-Object { get-filehash -path $_.FullName -Algorithm SHA256 } | format-list
- if: matrix.preset == 'CPU'
name: Verify Windows CPU payloads
shell: bash
run: |
set -euo pipefail
for payload in \
dist/windows-amd64/lib/ollama/llama-server.exe \
dist/windows-arm64/lib/ollama/llama-server.exe
do
[ -f "$payload" ] || { echo "missing $payload"; exit 1; }
done
- uses: actions/upload-artifact@v4
with:
name: depends-${{ matrix.os }}-${{ matrix.arch }}-${{ matrix.preset }}
path: dist\*
windows-build:
strategy:
matrix:
os: [windows]
arch: [amd64, arm64]
include:
- os: windows
arch: amd64
llvmarch: x86_64
- os: windows
arch: arm64
llvmarch: aarch64
runs-on: ${{ matrix.arch == 'arm64' && format('{0}-{1}', matrix.os, matrix.arch) || matrix.os }}
runs-on: windows
environment: release
needs: [setup-environment]
env:
GOFLAGS: ${{ needs.setup-environment.outputs.GOFLAGS }}
VERSION: ${{ needs.setup-environment.outputs.VERSION }}
steps:
- name: Install ARM64 system dependencies
if: matrix.arch == 'arm64'
run: |
$ErrorActionPreference = "Stop"
Set-ExecutionPolicy Bypass -Scope Process -Force
[System.Net.ServicePointManager]::SecurityProtocol = [System.Net.ServicePointManager]::SecurityProtocol -bor 3072
iex ((New-Object System.Net.WebClient).DownloadString('https://community.chocolatey.org/install.ps1'))
echo "C:\ProgramData\chocolatey\bin" | Out-File -FilePath $env:GITHUB_PATH -Encoding utf8 -Append
Invoke-WebRequest -Uri https://aka.ms/vs/17/release/vc_redist.arm64.exe -OutFile "${{ runner.temp }}\vc_redist.arm64.exe"
Start-Process -FilePath "${{ runner.temp }}\vc_redist.arm64.exe" -ArgumentList @("/install", "/quiet", "/norestart") -NoNewWindow -Wait
choco install -y --no-progress git gzip
echo "C:\Program Files\Git\cmd" | Out-File -FilePath $env:GITHUB_PATH -Encoding utf8 -Append
- name: Install clang and gcc-compat
run: |
$ErrorActionPreference = "Stop"
Set-ExecutionPolicy Bypass -Scope Process -Force
Invoke-WebRequest -Uri "https://github.com/mstorsjo/llvm-mingw/releases/download/20240619/llvm-mingw-20240619-ucrt-${{ matrix.llvmarch }}.zip" -OutFile "${{ runner.temp }}\llvm-mingw-ucrt.zip"
Invoke-WebRequest -Uri "https://github.com/mstorsjo/llvm-mingw/releases/download/20240619/llvm-mingw-20240619-ucrt-x86_64.zip" -OutFile "${{ runner.temp }}\llvm-mingw-ucrt.zip"
Expand-Archive -Path ${{ runner.temp }}\llvm-mingw-ucrt.zip -DestinationPath "C:\Program Files\"
$installPath=(Resolve-Path -Path "C:\Program Files\llvm-mingw-*-ucrt*").path
$installPath=(Resolve-Path -Path "C:\Program Files\llvm-mingw-*-ucrt-x86_64").path
echo "$installPath\bin" | Out-File -FilePath $env:GITHUB_PATH -Encoding utf8 -Append
if (!(Test-Path "$installPath\bin\aarch64-w64-mingw32-gcc.exe")) {
throw "llvm-mingw x86_64 package is missing the aarch64 cross compiler"
}
- uses: actions/checkout@v4
- uses: actions/setup-go@v5
with:
go-version-file: go.mod
cache-dependency-path: |
go.sum
Makefile.sync
LLAMA_CPP_VERSION
MLX_VERSION
MLX_C_VERSION
- name: Verify gcc is actually clang
run: |
$ErrorActionPreference='Continue'
@@ -323,20 +326,30 @@ jobs:
with:
node-version: "20"
- run: |
./scripts/build_windows ollama app
./scripts/build_windows ollama ollamaArm64 app appArm64
- name: Verify Windows build payloads
shell: bash
run: |
set -euo pipefail
for payload in \
dist/windows-amd64/ollama.exe \
dist/windows-arm64/ollama.exe
do
[ -f "$payload" ] || { echo "missing $payload"; exit 1; }
done
- name: Log build results
run: |
gci -path .\dist -Recurse -File | ForEach-Object { get-filehash -path $_.FullName -Algorithm SHA256 } | format-list
- uses: actions/upload-artifact@v4
with:
name: build-${{ matrix.os }}-${{ matrix.arch }}
name: build-windows-amd64
path: |
dist\*
windows-app:
runs-on: windows
environment: release
needs: [windows-build, windows-depends]
needs: [setup-environment, windows-build, windows-depends]
env:
GOFLAGS: ${{ needs.setup-environment.outputs.GOFLAGS }}
VERSION: ${{ needs.setup-environment.outputs.VERSION }}
@@ -362,7 +375,9 @@ jobs:
go-version-file: go.mod
cache-dependency-path: |
go.sum
Makefile.sync
LLAMA_CPP_VERSION
MLX_VERSION
MLX_C_VERSION
- uses: actions/download-artifact@v4
with:
pattern: depends-windows*
@@ -376,6 +391,18 @@ jobs:
- name: Log dist contents after download
run: |
gci -path .\dist -recurse
- name: Verify Windows package inputs
shell: bash
run: |
set -euo pipefail
for payload in \
dist/windows-amd64/ollama.exe \
dist/windows-amd64/lib/ollama/llama-server.exe \
dist/windows-arm64/ollama.exe \
dist/windows-arm64/lib/ollama/llama-server.exe
do
[ -f "$payload" ] || { echo "missing $payload"; exit 1; }
done
- run: |
./scripts/build_windows.ps1 deps sign installer zip
- name: Log contents after build
@@ -389,31 +416,28 @@ jobs:
dist/*.ps1
dist/OllamaSetup.exe
# Pre-build each Dockerfile stage on its own runner in parallel and push the
# resulting layers to a per-stage registry cache. The downstream
# docker-build-push job then assembles cache-hit-only.
linux-depends:
strategy:
matrix:
include:
- arch: amd64
target: cpu
target: llama-server-cpu
- arch: amd64
target: cuda-12
target: llama-server-cuda_v12
- arch: amd64
target: cuda-13
target: llama-server-cuda_v13
- arch: amd64
target: mlx
- arch: amd64
target: rocm-7
target: llama-server-rocm_v7_2
- arch: amd64
target: vulkan
target: llama-server-vulkan
- arch: arm64
target: cpu
target: llama-server-cpu
- arch: arm64
target: cuda-12
target: llama-server-cuda_v12
- arch: arm64
target: cuda-13
target: llama-server-cuda_v13
- arch: arm64
target: jetpack-5
- arch: arm64
@@ -430,7 +454,6 @@ jobs:
with:
username: ${{ vars.DOCKER_USER }}
password: ${{ secrets.DOCKER_ACCESS_TOKEN }}
# Increase swap to handle momentary spikes in RAM from NVCC compiles
- if: matrix.target == 'mlx'
name: Increase Linux swap to 200 GB
shell: bash
@@ -459,12 +482,13 @@ jobs:
provenance: false
sbom: false
build-args: |
GOFLAGS=${{ env.GOFLAGS }}
CGO_CFLAGS=${{ env.CGO_CFLAGS }}
CGO_CXXFLAGS=${{ env.CGO_CXXFLAGS }}
GOFLAGS=${{ env.GOFLAGS }}
APT_MIRROR=http://azure.archive.ubuntu.com/ubuntu
OLLAMA_MLX_BUILD_JOBS=16
OLLAMA_MLX_NVCC_THREADS=6
APT_MIRROR=http://azure.archive.ubuntu.com/ubuntu
APT_PORTS_MIRROR=http://azure.ports.ubuntu.com/ubuntu-ports
cache-from: |
type=registry,ref=ollama/release:cache-${{ matrix.arch }}-${{ matrix.target }}
type=registry,ref=${{ vars.DOCKER_REPO }}:latest
@@ -472,58 +496,65 @@ jobs:
# Build each Docker variant (OS, arch, and flavor) separately. Using QEMU is unreliable and slower.
# Heavy stages were pre-built by linux-depends; this job is cache-hit-only for those layers
# and just assembles, runs the Go build, and pushes the final image.
# and just assembles, runs the Go build, pushes the final image, and extracts release bundles.
docker-build-push:
strategy:
matrix:
include:
- os: linux
arch: arm64
archive-target: archive
build-args: |
CGO_CFLAGS
CGO_CXXFLAGS
GOFLAGS
APT_MIRROR=http://azure.archive.ubuntu.com/ubuntu
APT_PORTS_MIRROR=http://azure.ports.ubuntu.com/ubuntu-ports
OLLAMA_MLX_BUILD_JOBS=16
OLLAMA_MLX_NVCC_THREADS=6
cache-from: |
type=registry,ref=${{ vars.DOCKER_REPO }}:latest
type=registry,ref=ollama/release:cache-arm64-cpu
type=registry,ref=ollama/release:cache-arm64-cuda-12
type=registry,ref=ollama/release:cache-arm64-cuda-13
type=registry,ref=ollama/release:cache-arm64-llama-server-cpu
type=registry,ref=ollama/release:cache-arm64-llama-server-cuda_v12
type=registry,ref=ollama/release:cache-arm64-llama-server-cuda_v13
type=registry,ref=ollama/release:cache-arm64-jetpack-5
type=registry,ref=ollama/release:cache-arm64-jetpack-6
type=registry,ref=${{ vars.DOCKER_REPO }}:latest
- os: linux
arch: amd64
archive-target: archive
build-args: |
CGO_CFLAGS
CGO_CXXFLAGS
GOFLAGS
APT_MIRROR=http://azure.archive.ubuntu.com/ubuntu
APT_PORTS_MIRROR=http://azure.ports.ubuntu.com/ubuntu-ports
OLLAMA_MLX_BUILD_JOBS=16
OLLAMA_MLX_NVCC_THREADS=6
cache-from: |
type=registry,ref=${{ vars.DOCKER_REPO }}:latest
type=registry,ref=ollama/release:cache-amd64-cpu
type=registry,ref=ollama/release:cache-amd64-cuda-12
type=registry,ref=ollama/release:cache-amd64-cuda-13
type=registry,ref=ollama/release:cache-amd64-llama-server-cpu
type=registry,ref=ollama/release:cache-amd64-llama-server-cuda_v12
type=registry,ref=ollama/release:cache-amd64-llama-server-cuda_v13
type=registry,ref=ollama/release:cache-amd64-mlx
type=registry,ref=ollama/release:cache-amd64-vulkan
type=registry,ref=ollama/release:cache-amd64-llama-server-rocm_v7_2
type=registry,ref=ollama/release:cache-amd64-llama-server-vulkan
type=registry,ref=${{ vars.DOCKER_REPO }}:latest
- os: linux
arch: amd64
suffix: '-rocm'
archive-target: image-archive
build-args: |
CGO_CFLAGS
CGO_CXXFLAGS
GOFLAGS
FLAVOR=rocm
APT_MIRROR=http://azure.archive.ubuntu.com/ubuntu
APT_PORTS_MIRROR=http://azure.ports.ubuntu.com/ubuntu-ports
OLLAMA_MLX_BUILD_JOBS=16
OLLAMA_MLX_NVCC_THREADS=6
cache-from: |
type=registry,ref=ollama/release:cache-amd64-llama-server-cpu
type=registry,ref=ollama/release:cache-amd64-llama-server-rocm_v7_2
type=registry,ref=${{ vars.DOCKER_REPO }}:latest
type=registry,ref=ollama/release:cache-amd64-cpu
type=registry,ref=ollama/release:cache-amd64-rocm-7
runs-on: ${{ matrix.arch == 'arm64' && format('{0}-{1}', matrix.os, matrix.arch) || matrix.os }}
environment: release
needs: [setup-environment, linux-depends]
@@ -556,14 +587,11 @@ jobs:
name: digest-${{ matrix.os }}-${{ matrix.arch }}-${{ matrix.suffix }}
path: |
${{ runner.temp }}/${{ matrix.os }}-${{ matrix.arch }}-${{ matrix.suffix }}.txt
# Re-run buildx with --target archive against buildkit's local cache to
# extract the release directory layout. All upstream stages were just
# built above, so this is a cache-hit-only pass that just writes files.
- uses: docker/build-push-action@v6
with:
context: .
platforms: ${{ matrix.os }}/${{ matrix.arch }}
target: archive
target: ${{ matrix.archive-target }}
provenance: false
sbom: false
build-args: ${{ matrix.build-args }}
@@ -572,24 +600,33 @@ jobs:
- name: Deduplicate CUDA libraries
run: |
./scripts/deduplicate_cuda_libs.sh dist/${{ matrix.os }}-${{ matrix.arch }}
- name: Verify Linux build payloads
shell: bash
run: |
set -euo pipefail
base="dist/${{ matrix.os }}-${{ matrix.arch }}"
for payload in \
"$base/bin/ollama" \
"$base/lib/ollama/llama-server"
do
[ -f "$payload" ] || { echo "missing $payload"; exit 1; }
done
- run: |
for COMPONENT in bin/* lib/ollama/*; do
case "$COMPONENT" in
bin/ollama*) echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}.tar.in ;;
lib/ollama/*.so*) echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}.tar.in ;;
lib/ollama/llama-server*|lib/ollama/llama-quantize*) echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}.tar.in ;;
lib/ollama/cuda_v*) echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}.tar.in ;;
lib/ollama/vulkan*) echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}.tar.in ;;
lib/ollama/mlx*) echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}-mlx.tar.in ;;
lib/ollama/include*) echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}.tar.in ;;
lib/ollama/include*) echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}-mlx.tar.in ;;
lib/ollama/cuda_jetpack5) echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}-jetpack5.tar.in ;;
lib/ollama/cuda_jetpack6) echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}-jetpack6.tar.in ;;
lib/ollama/rocm) echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}-rocm.tar.in ;;
lib/ollama/rocm_v*) echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}-rocm.tar.in ;;
esac
done
working-directory: dist/${{ matrix.os }}-${{ matrix.arch }}
# rocm builds cpu + rocm libs for the container image, which
# creates a CPU-only amd64 tarball that would collide with the full
# bundle when the release job merges artifacts.
- if: matrix.suffix == '-rocm'
run: rm -f dist/${{ matrix.os }}-${{ matrix.arch }}/ollama-${{ matrix.os }}-${{ matrix.arch }}.tar.in
- run: |
@@ -665,6 +702,21 @@ jobs:
- name: Copy install scripts to dist
run: |
cp scripts/install.sh dist/install.sh
- name: Verify release artifacts
run: |
required=(
dist/OllamaSetup.exe
dist/install.ps1
dist/install.sh
dist/ollama-windows-amd64.zip
dist/ollama-windows-arm64.zip
)
for payload in "${required[@]}"; do
if [ ! -f "$payload" ]; then
echo "::error::Missing expected release artifact: $payload"
exit 1
fi
done
- name: Generate checksum file
run: find . -type f -not -name 'sha256sum.txt' | xargs sha256sum | tee sha256sum.txt
working-directory: dist

189
.github/workflows/test.yaml vendored
View File

@@ -23,7 +23,7 @@ jobs:
outputs:
changed: ${{ steps.changes.outputs.changed }}
app_changed: ${{ steps.changes.outputs.app_changed }}
vendorsha: ${{ steps.changes.outputs.vendorsha }}
enginehash: ${{ steps.changes.outputs.enginehash }}
steps:
- uses: actions/checkout@v4
with:
@@ -38,9 +38,42 @@ jobs:
| xargs python3 -c "import sys; from pathlib import Path; print(any(Path(x).match(glob) for x in sys.argv[1:] for glob in '$*'.split(' ')))"
}
echo changed=$(changed 'llama/llama.cpp/**/*' 'ml/backend/ggml/ggml/**/*' '.github/**/*') | tee -a $GITHUB_OUTPUT
echo changed=$(changed \
'CMakeLists.txt' \
'CMakePresets.json' \
'cmake/**' \
'cmake/**/*' \
'llama/server/**/*' \
'llama/compat/**/*' \
'LLAMA_CPP_VERSION' \
'MLX_VERSION' \
'MLX_C_VERSION' \
'llama/llama.cpp/**/*' \
'ml/backend/ggml/ggml/**/*' \
'x/imagegen/mlx/**' \
'x/imagegen/mlx/**/*' \
'.github/**/*') | tee -a $GITHUB_OUTPUT
echo app_changed=$(changed 'app/**' 'app/**/*') | tee -a $GITHUB_OUTPUT
echo vendorsha=$(make -f Makefile.sync print-base) | tee -a $GITHUB_OUTPUT
echo enginehash=$(cat LLAMA_CPP_VERSION)-$(cat MLX_VERSION)-$(cat MLX_C_VERSION) | tee -a $GITHUB_OUTPUT
patches:
strategy:
matrix:
os: [ubuntu-latest, windows-latest]
runs-on: ${{ matrix.os }}
steps:
- uses: actions/checkout@v4
- name: Verify patches apply cleanly
shell: bash
run: |
cmake -S llama/server -B "$RUNNER_TEMP/llama-server-patch-check" \
-DCMAKE_BUILD_TYPE=Release \
-DBUILD_SHARED_LIBS=ON \
-DGGML_BACKEND_DL=ON \
-DGGML_NATIVE=OFF \
-DGGML_OPENMP=OFF \
-DGGML_CPU_ALL_VARIANTS=ON \
-DOLLAMA_RUNNER_DIR=
linux:
needs: [changes]
@@ -49,23 +82,41 @@ jobs:
matrix:
include:
- preset: CPU
superbuild_target: ollama-local
superbuild_dir: build/local-superbuild
superbuild_args: ''
expected_payload: lib/ollama/llama-server
install-go: true
- preset: CUDA
container: nvidia/cuda:13.0.0-devel-ubuntu22.04
flags: '-DCMAKE_CUDA_ARCHITECTURES=87'
superbuild_target: ollama-llama-server-cuda_v13
superbuild_dir: build/local-superbuild-cuda_v13
superbuild_args: '-DOLLAMA_LLAMA_BACKENDS=cuda_v13 -DCMAKE_CUDA_ARCHITECTURES=87'
expected_payload: lib/ollama/cuda_v13/libggml-cuda.so
- preset: ROCm
container: rocm/dev-ubuntu-22.04:7.2.1
extra-packages: rocm-libs
flags: '-DAMDGPU_TARGETS=gfx1010 -DCMAKE_PREFIX_PATH=/opt/rocm'
superbuild_target: ollama-llama-server-rocm_v7_2
superbuild_dir: build/local-superbuild-rocm_v7_2
superbuild_args: '-DOLLAMA_LLAMA_BACKENDS=rocm_v7_2 -DAMDGPU_TARGETS=gfx1010 -DCMAKE_PREFIX_PATH=/opt/rocm'
expected_payload: lib/ollama/rocm_v7_2/libggml-hip.so
- preset: Vulkan
container: ubuntu:22.04
extra-packages: >
mesa-vulkan-drivers vulkan-tools
libvulkan1 libvulkan-dev
vulkan-sdk cmake ccache g++ make
vulkan-sdk spirv-headers cmake ccache g++ make
superbuild_target: ollama-llama-server-vulkan
superbuild_dir: build/local-superbuild-vulkan
superbuild_args: '-DOLLAMA_LLAMA_BACKENDS=vulkan'
expected_payload: lib/ollama/vulkan/libggml-vulkan.so
- preset: 'MLX CUDA 13'
container: nvidia/cuda:13.0.0-devel-ubuntu22.04
extra-packages: libcudnn9-dev-cuda-13 libopenblas-dev liblapack-dev liblapacke-dev git curl
flags: '-DCMAKE_CUDA_ARCHITECTURES=87 -DMLX_CUDA_ARCHITECTURES=80-virtual -DBLAS_INCLUDE_DIRS=/usr/include/x86_64-linux-gnu -DLAPACK_INCLUDE_DIRS=/usr/include/x86_64-linux-gnu'
superbuild_target: ollama-mlx-cuda_v13
superbuild_dir: build/local-superbuild-mlx-cuda_v13
superbuild_args: '-DOLLAMA_MLX_BACKENDS=cuda_v13 -DCMAKE_CUDA_ARCHITECTURES=87 -DMLX_CUDA_ARCHITECTURES=80-virtual -DBLAS_INCLUDE_DIRS=/usr/include/x86_64-linux-gnu -DLAPACK_INCLUDE_DIRS=/usr/include/x86_64-linux-gnu'
expected_payload: lib/ollama/mlx_cuda_v13/libmlx.so
install-go: true
runs-on: linux
container: ${{ matrix.container }}
@@ -82,11 +133,9 @@ jobs:
echo "deb [signed-by=/usr/share/keyrings/lunarg-archive-keyring.gpg] https://packages.lunarg.com/vulkan/1.4.313 jammy main" | $sudo tee /etc/apt/sources.list.d/lunarg-vulkan-1.4.313-jammy.list > /dev/null
$sudo apt-get update
fi
$sudo apt-get install -y cmake ccache ${{ matrix.extra-packages }}
# MLX requires CMake 3.25+, install from official releases
if [ "${{ matrix.preset }}" = "MLX CUDA 13" ]; then
curl -fsSL https://github.com/Kitware/CMake/releases/download/v3.31.2/cmake-3.31.2-linux-$(uname -m).tar.gz | $sudo tar xz -C /usr/local --strip-components 1
fi
$sudo apt-get install -y cmake ccache curl git ${{ matrix.extra-packages }}
# Use a current CMake for upstream llama.cpp and Vulkan dependency discovery.
curl -fsSL https://github.com/Kitware/CMake/releases/download/v3.31.2/cmake-3.31.2-linux-$(uname -m).tar.gz | $sudo tar xz -C /usr/local --strip-components 1
# Export VULKAN_SDK if provided by LunarG package (defensive)
if [ -d "/usr/lib/x86_64-linux-gnu/vulkan" ] && [ "${{ matrix.preset }}" = "Vulkan" ]; then
echo "VULKAN_SDK=/usr" >> $GITHUB_ENV
@@ -96,17 +145,30 @@ jobs:
- if: matrix.install-go
name: Install Go
run: |
[ -n "${{ matrix.container }}" ] || sudo=sudo
GO_VERSION=$(awk '/^go / { print $2 }' go.mod)
curl -fsSL "https://golang.org/dl/go${GO_VERSION}.linux-$(dpkg --print-architecture).tar.gz" | tar xz -C /usr/local
curl -fsSL "https://golang.org/dl/go${GO_VERSION}.linux-$(dpkg --print-architecture).tar.gz" | $sudo tar xz -C /usr/local
echo "/usr/local/go/bin" >> $GITHUB_PATH
- uses: actions/cache@v4
with:
path: /github/home/.cache/ccache
key: ccache-${{ runner.os }}-${{ runner.arch }}-${{ matrix.preset }}-${{ needs.changes.outputs.vendorsha }}
- run: |
cmake --preset "${{ matrix.preset }}" ${{ matrix.flags }}
cmake --build --preset "${{ matrix.preset }}" -- -l $(nproc)
key: ccache-${{ runner.os }}-${{ runner.arch }}-${{ matrix.preset }}-${{ needs.changes.outputs.enginehash }}
- name: Build native superbuild
if: matrix.superbuild_target
run: |
cmake -S . -B "${{ matrix.superbuild_dir }}" ${{ matrix.superbuild_args }}
CMAKE_BUILD_PARALLEL_LEVEL=$(nproc) cmake --build "${{ matrix.superbuild_dir }}" --target "${{ matrix.superbuild_target }}" -- -l $(nproc)
test -e "${{ matrix.superbuild_dir }}/${{ matrix.expected_payload }}"
- name: Verify local superbuild install
if: matrix.superbuild_target == 'ollama-local'
run: |
./ollama --version
"${{ matrix.superbuild_dir }}/lib/ollama/llama-server" --version
test -x "${{ matrix.superbuild_dir }}/lib/ollama/llama-quantize"
cmake --install "${{ matrix.superbuild_dir }}" --component ollama-local --prefix "$RUNNER_TEMP/ollama-local"
"$RUNNER_TEMP/ollama-local/bin/ollama" --version
"$RUNNER_TEMP/ollama-local/lib/ollama/llama-server" --version
test -x "$RUNNER_TEMP/ollama-local/lib/ollama/llama-quantize"
windows:
needs: [changes]
if: needs.changes.outputs.changed == 'True'
@@ -114,9 +176,16 @@ jobs:
matrix:
include:
- preset: CPU
superbuild_target: ollama-local
superbuild_dir: build\local-superbuild
superbuild_args: ''
expected_payload: lib\ollama\llama-server.exe
- preset: CUDA
install: https://developer.download.nvidia.com/compute/cuda/13.0.0/local_installers/cuda_13.0.0_windows.exe
flags: '-DCMAKE_CUDA_ARCHITECTURES=80'
superbuild_target: ollama-llama-server-cuda_v13
superbuild_dir: build\local-superbuild-cuda_v13
superbuild_args: '-DOLLAMA_LLAMA_BACKENDS=cuda_v13 -DCMAKE_CUDA_ARCHITECTURES=80'
expected_payload: lib\ollama\cuda_v13\ggml-cuda.dll
cuda-components:
- '"cudart"'
- '"nvcc"'
@@ -127,14 +196,26 @@ jobs:
- '"nvptxcompiler"'
cuda-version: '13.0'
- preset: ROCm
install: https://download.amd.com/developer/eula/rocm-hub/AMD-Software-PRO-Edition-24.Q4-WinSvr2022-For-HIP.exe
flags: '-DAMDGPU_TARGETS=gfx1010 -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DCMAKE_C_FLAGS="-parallel-jobs=4 -Wno-ignored-attributes -Wno-deprecated-pragma" -DCMAKE_CXX_FLAGS="-parallel-jobs=4 -Wno-ignored-attributes -Wno-deprecated-pragma"'
install: https://download.amd.com/developer/eula/rocm-hub/AMD-Software-PRO-Edition-26.Q1-Win11-For-HIP.exe
rocm-version: '7.1'
superbuild_target: ollama-llama-server-rocm_v7_1
superbuild_dir: build\local-superbuild-rocm_v7_1
superbuild_args: '-DOLLAMA_LLAMA_BACKENDS=rocm_v7_1 -DAMDGPU_TARGETS=gfx1010'
expected_payload: lib\ollama\rocm_v7_1\ggml-hip.dll
- preset: Vulkan
install: https://sdk.lunarg.com/sdk/download/1.4.321.1/windows/vulkansdk-windows-X64-1.4.321.1.exe
superbuild_target: ollama-llama-server-vulkan
superbuild_dir: build\local-superbuild-vulkan
superbuild_args: '-DOLLAMA_LLAMA_BACKENDS=vulkan'
expected_payload: lib\ollama\vulkan\ggml-vulkan.dll
- preset: 'MLX CUDA 13'
install: https://developer.download.nvidia.com/compute/cuda/13.0.0/local_installers/cuda_13.0.0_windows.exe
cudnn-install: https://developer.download.nvidia.com/compute/cudnn/redist/cudnn/windows-x86_64/cudnn-windows-x86_64-9.18.1.3_cuda13-archive.zip
flags: '-DCMAKE_CUDA_ARCHITECTURES=80 -DMLX_CUDA_ARCHITECTURES=80-virtual'
superbuild_target: ollama-mlx-cuda_v13
superbuild_dir: build\local-superbuild-mlx-cuda_v13
superbuild_args: '-DOLLAMA_MLX_BACKENDS=cuda_v13 -DCMAKE_CUDA_ARCHITECTURES=80 -DMLX_CUDA_ARCHITECTURES=80-virtual'
expected_payload: lib\ollama\mlx_cuda_v13\mlx.dll
install-go: true
cuda-components:
- '"cudart"'
- '"nvcc"'
@@ -203,6 +284,10 @@ jobs:
}
$vulkanPath = (Resolve-Path "C:\VulkanSDK\*").path
$vulkanRuntime = Join-Path $vulkanPath "Helpers\VulkanRT.exe"
if (Test-Path $vulkanRuntime) {
Start-Process -FilePath $vulkanRuntime -ArgumentList "/s" -NoNewWindow -Wait
}
echo "$vulkanPath\bin" | Out-File -FilePath $env:GITHUB_PATH -Encoding utf8 -Append
echo "VULKAN_SDK=$vulkanPath" >> $env:GITHUB_ENV
- if: matrix.preset == 'MLX CUDA 13'
@@ -232,18 +317,44 @@ jobs:
C:\Program Files\NVIDIA\CUDNN
key: ${{ matrix.install }}-${{ matrix.cudnn-install }}
- uses: actions/checkout@v4
- if: matrix.superbuild_target == 'ollama-local' || matrix.install-go
uses: actions/setup-go@v5
with:
go-version-file: 'go.mod'
- uses: actions/cache@v4
with:
path: ${{ github.workspace }}\.ccache
key: ccache-${{ runner.os }}-${{ runner.arch }}-${{ matrix.preset }}-${{ needs.changes.outputs.vendorsha }}
- run: |
key: ccache-${{ runner.os }}-${{ runner.arch }}-${{ matrix.preset }}-${{ needs.changes.outputs.enginehash }}
- name: Build native superbuild
if: matrix.superbuild_target
run: |
$ErrorActionPreference = "Stop"
Import-Module 'C:\Program Files\Microsoft Visual Studio\2022\Enterprise\Common7\Tools\Microsoft.VisualStudio.DevShell.dll'
Enter-VsDevShell -VsInstallPath 'C:\Program Files\Microsoft Visual Studio\2022\Enterprise' -SkipAutomaticLocation -DevCmdArguments '-arch=x64 -no_logo'
cmake --preset "${{ matrix.preset }}" ${{ matrix.flags }}
cmake --build --preset "${{ matrix.preset }}" -- -l $([Environment]::ProcessorCount)
cmake -S . -B "${{ matrix.superbuild_dir }}" ${{ matrix.superbuild_args }}
$env:CMAKE_BUILD_PARALLEL_LEVEL = [Environment]::ProcessorCount
cmake --build "${{ matrix.superbuild_dir }}" --target "${{ matrix.superbuild_target }}" -- -l $([Environment]::ProcessorCount)
if (!(Test-Path "${{ matrix.superbuild_dir }}\${{ matrix.expected_payload }}")) {
throw "missing ${{ matrix.expected_payload }}"
}
env:
CMAKE_GENERATOR: Ninja
- name: Verify local superbuild install
if: matrix.superbuild_target == 'ollama-local'
run: |
$ErrorActionPreference = "Stop"
& ".\ollama.exe" --version
& "${{ matrix.superbuild_dir }}\lib\ollama\llama-server.exe" --version
if (!(Test-Path "${{ matrix.superbuild_dir }}\lib\ollama\llama-quantize.exe")) {
throw "missing llama-quantize.exe"
}
$installPrefix = Join-Path $env:RUNNER_TEMP "ollama-local"
cmake --install "${{ matrix.superbuild_dir }}" --component ollama-local --prefix "$installPrefix"
& "$installPrefix\bin\ollama.exe" --version
& "$installPrefix\lib\ollama\llama-server.exe" --version
if (!(Test-Path "$installPrefix\lib\ollama\llama-quantize.exe")) {
throw "missing installed llama-quantize.exe"
}
go_mod_tidy:
runs-on: ubuntu-latest
steps:
@@ -266,7 +377,9 @@ jobs:
go-version-file: 'go.mod'
cache-dependency-path: |
go.sum
Makefile.sync
LLAMA_CPP_VERSION
MLX_VERSION
MLX_C_VERSION
- uses: actions/setup-node@v4
with:
node-version: '20'
@@ -280,6 +393,17 @@ jobs:
if: ${{ startsWith(matrix.os, 'ubuntu') }}
working-directory: ./app/ui/app
run: npm test
- name: Verify MLX generated files are current
if: ${{ startsWith(matrix.os, 'ubuntu') }}
run: |
cmake -S . -B build/mlx-generate -DOLLAMA_MLX_BACKENDS=cuda_v13
cmake --build build/mlx-generate --target ollama-mlx-generate-wrappers
git diff --exit-code -- \
x/imagegen/mlx/mlx.h \
x/imagegen/mlx/mlx.c \
x/mlxrunner/mlx/generated.h \
x/mlxrunner/mlx/generated.c \
x/mlxrunner/mlx/include/mlx/c
- name: Run go generate
run: go generate ./...
@@ -294,12 +418,3 @@ jobs:
- uses: golangci/golangci-lint-action@v9
with:
only-new-issues: true
patches:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- name: Verify patches apply cleanly and do not change files
run: |
make -f Makefile.sync clean checkout apply-patches sync
git diff --compact-summary --exit-code