Add support for new models and fix GitHub issues

- Add Gemma3n model support with text generation capabilities
- Add new CUDA mean operations for improved performance
- Add macOS documentation and performance tests
- Update LLAMA patches for ROCm/CUDA compatibility
- Fix various model conversion and processing issues
- Update CI workflows and build configurations
- Add library model tests and Shakespeare test data

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
Shang Chieh Tseng
2025-07-20 00:12:36 +08:00
parent 1fa71c2670
commit cbcbc9ae07
81 changed files with 132316 additions and 747 deletions

View File

@@ -23,7 +23,7 @@ jobs:
echo GOFLAGS="'-ldflags=-w -s \"-X=github.com/ollama/ollama/version.Version=${GITHUB_REF_NAME#v}\" \"-X=github.com/ollama/ollama/server.mode=release\"'" >>$GITHUB_OUTPUT
darwin-build:
runs-on: macos-13
runs-on: macos-13-xlarge
environment: release
needs: setup-environment
strategy:
@@ -54,48 +54,6 @@ jobs:
name: build-${{ matrix.os }}-${{ matrix.arch }}
path: dist/*
darwin-sign:
runs-on: macos-13
environment: release
needs: darwin-build
steps:
- uses: actions/checkout@v4
- run: |
echo $MACOS_SIGNING_KEY | base64 --decode > certificate.p12
security create-keychain -p password build.keychain
security default-keychain -s build.keychain
security unlock-keychain -p password build.keychain
security import certificate.p12 -k build.keychain -P $MACOS_SIGNING_KEY_PASSWORD -T /usr/bin/codesign
security set-key-partition-list -S apple-tool:,apple:,codesign: -s -k password build.keychain
security set-keychain-settings -lut 3600 build.keychain
env:
MACOS_SIGNING_KEY: ${{ secrets.MACOS_SIGNING_KEY }}
MACOS_SIGNING_KEY_PASSWORD: ${{ secrets.MACOS_SIGNING_KEY_PASSWORD }}
- uses: actions/download-artifact@v4
with:
name: build-darwin-amd64
path: dist/darwin-amd64
- uses: actions/download-artifact@v4
with:
name: build-darwin-arm64
path: dist/darwin-arm64
- run: |
export VERSION=${GITHUB_REF_NAME#v}
./scripts/build_darwin.sh sign macapp
env:
APPLE_IDENTITY: ${{ secrets.APPLE_IDENTITY }}
APPLE_PASSWORD: ${{ secrets.APPLE_PASSWORD }}
APPLE_TEAM_ID: ${{ vars.APPLE_TEAM_ID }}
APPLE_ID: ${{ vars.APPLE_ID }}
SDKROOT: /Applications/Xcode_14.1.0.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk
DEVELOPER_DIR: /Applications/Xcode_14.1.0.app/Contents/Developer
- uses: actions/upload-artifact@v4
with:
name: dist-darwin
path: |
dist/Ollama-darwin.zip
dist/ollama-darwin.tgz
windows-depends:
strategy:
matrix:
@@ -103,21 +61,18 @@ jobs:
arch: [amd64]
preset: ['CPU']
include:
- os: windows
arch: amd64
preset: 'CUDA 11'
install: https://developer.download.nvidia.com/compute/cuda/11.3.1/local_installers/cuda_11.3.1_465.89_win10.exe
cuda-version: '11.3'
- os: windows
arch: amd64
preset: 'CUDA 12'
install: https://developer.download.nvidia.com/compute/cuda/12.8.0/local_installers/cuda_12.8.0_571.96_windows.exe
cuda-version: '12.8'
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"'
runs-on: ${{ matrix.arch == 'arm64' && format('{0}-{1}', matrix.os, matrix.arch) || matrix.os }}
environment: release
env:
@@ -160,6 +115,9 @@ jobs:
echo "$hipPath\bin" | Out-File -FilePath $env:GITHUB_PATH -Encoding utf8 -Append
echo "CC=$hipPath\bin\clang.exe" | Out-File -FilePath $env:GITHUB_ENV -Append
echo "CXX=$hipPath\bin\clang++.exe" | Out-File -FilePath $env:GITHUB_ENV -Append
echo "HIPCXX=$hipPath\bin\clang++.exe" | Out-File -FilePath $env:GITHUB_ENV -Append
echo "HIP_PLATFORM=amd" | Out-File -FilePath $env:GITHUB_ENV -Append
echo "CMAKE_PREFIX_PATH=$hipPath" | Out-File -FilePath $env:GITHUB_ENV -Append
- if: matrix.preset == 'CPU'
run: |
echo "CC=clang.exe" | Out-File -FilePath $env:GITHUB_ENV -Append
@@ -178,9 +136,9 @@ jobs:
key: ccache-${{ matrix.os }}-${{ matrix.arch }}-${{ matrix.preset }}
- name: Build target "${{ matrix.preset }}"
run: |
Import-Module 'C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\Common7\Tools\Microsoft.VisualStudio.DevShell.dll'
Enter-VsDevShell -VsInstallPath 'C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise' -SkipAutomaticLocation -DevCmdArguments '-arch=x64 -no_logo'
cmake --preset "${{ matrix.preset }}"
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 }}" ${{ matrix.flags }}
cmake --build --parallel --preset "${{ matrix.preset }}"
cmake --install build --component "${{ startsWith(matrix.preset, 'CUDA ') && 'CUDA' || startsWith(matrix.preset, 'ROCm ') && 'HIP' || 'CPU' }}" --strip --parallel 8
env:
@@ -230,61 +188,11 @@ jobs:
go-version-file: go.mod
- run: |
go build -o dist/${{ matrix.os }}-${{ matrix.arch }}/ .
- if: matrix.arch == 'arm64'
run: |
Invoke-WebRequest -Uri "https://aka.ms/vs/17/release/vc_redist.arm64.exe" -OutFile "dist\windows-arm64\vc_redist.arm64.exe"
- run: |
$env:VERSION='${{ github.ref_name }}' -Replace "v(.*)", '$1'
& .\scripts\build_windows.ps1 buildApp
env:
VCToolsRedistDir: stub
- uses: actions/upload-artifact@v4
with:
name: build-${{ matrix.os }}-${{ matrix.arch }}
path: |
dist\${{ matrix.os }}-${{ matrix.arch }}\*.exe
dist\${{ matrix.os }}-${{ matrix.arch }}-app.exe
windows-sign:
runs-on: windows-2022
environment: release
needs: [windows-depends, windows-build]
steps:
- uses: actions/checkout@v4
- uses: google-github-actions/auth@v2
with:
project_id: ollama
credentials_json: ${{ secrets.GOOGLE_SIGNING_CREDENTIALS }}
- run: |
$ErrorActionPreference = "Stop"
Invoke-WebRequest -Uri "https://go.microsoft.com/fwlink/p/?LinkId=323507" -OutFile "${{ runner.temp }}\sdksetup.exe"
Start-Process "${{ runner.temp }}\sdksetup.exe" -ArgumentList @("/q") -NoNewWindow -Wait
Invoke-WebRequest -Uri "https://github.com/GoogleCloudPlatform/kms-integrations/releases/download/cng-v1.0/kmscng-1.0-windows-amd64.zip" -OutFile "${{ runner.temp }}\plugin.zip"
Expand-Archive -Path "${{ runner.temp }}\plugin.zip" -DestinationPath "${{ runner.temp }}\plugin\"
& "${{ runner.temp }}\plugin\*\kmscng.msi" /quiet
echo "${{ vars.OLLAMA_CERT }}" >ollama_inc.crt
- uses: actions/download-artifact@v4
with:
pattern: build-windows-*
path: dist\
merge-multiple: true
- uses: actions/download-artifact@v4
with:
pattern: depends-windows-amd64-*
path: dist\windows-amd64\
merge-multiple: true
- run: |
& .\scripts\build_windows.ps1 gatherDependencies sign buildInstaller distZip
env:
KEY_CONTAINER: ${{ vars.KEY_CONTAINER }}
- uses: actions/upload-artifact@v4
with:
name: dist-windows
path: |
dist\OllamaSetup.exe
dist\ollama-windows-*.zip
linux-build:
strategy:
@@ -317,21 +225,26 @@ jobs:
CGO_CFLAGS=${{ env.CGO_CFLAGS }}
CGO_CXXFLAGS=${{ env.CGO_CXXFLAGS }}
outputs: type=local,dest=dist/${{ matrix.os }}-${{ matrix.arch }}
cache-from: type=registry,ref=ollama/ollama:latest
cache-from: type=registry,ref=${{ vars.DOCKER_REPO }}:latest
cache-to: type=inline
- 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/cuda_v11) echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}.tar.in ;;
lib/ollama/cuda_v12) echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}.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 ;;
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/cuda_sbsa) echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}.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 }}-${{ matrix.arch }}-rocm.tar.in ;;
esac
done
working-directory: dist/${{ matrix.os }}-${{ matrix.arch }}
- run: |
echo "Manifests"
for ARCHIVE in dist/${{ matrix.os }}-${{ matrix.arch }}/*.tar.in ; do
echo $ARCHIVE
cat $ARCHIVE
done
- run: |
for ARCHIVE in dist/${{ matrix.os }}-${{ matrix.arch }}/*.tar.in; do
tar c -C dist/${{ matrix.os }}-${{ matrix.arch }} -T $ARCHIVE --owner 0 --group 0 | pigz -9vc >$(basename ${ARCHIVE//.*/}.tgz);
@@ -385,8 +298,8 @@ jobs:
context: .
platforms: ${{ matrix.os }}/${{ matrix.arch }}
build-args: ${{ matrix.build-args }}
outputs: type=image,name=ollama/ollama,push-by-digest=true,name-canonical=true,push=true
cache-from: type=registry,ref=ollama/ollama:latest
outputs: type=image,name=${{ vars.DOCKER_REPO }},push-by-digest=true,name-canonical=true,push=true
cache-from: type=registry,ref=${{ vars.DOCKER_REPO }}:latest
cache-to: type=inline
- run: |
mkdir -p ${{ matrix.os }}-${{ matrix.arch }}
@@ -418,7 +331,7 @@ jobs:
latest=false
suffix=${{ matrix.suffix }}
images: |
ollama/ollama
${{ vars.DOCKER_REPO }}
tags: |
type=ref,enable=true,priority=600,prefix=pr-,event=pr
type=semver,pattern={{version}}
@@ -428,56 +341,24 @@ jobs:
path: ${{ runner.temp }}
merge-multiple: true
- run: |
docker buildx imagetools create $(echo '${{ steps.metadata.outputs.json }}' | jq -cr '.tags | map("-t", .) | join(" ")') $(cat *-${{ matrix.suffix }}.txt | xargs printf 'ollama/ollama@%s ')
docker buildx imagetools inspect ollama/ollama:${{ steps.metadata.outputs.version }}
docker buildx imagetools create $(echo '${{ steps.metadata.outputs.json }}' | jq -cr '.tags | map("-t", .) | join(" ")') $(cat *-${{ matrix.suffix }}.txt | xargs printf '${{ vars.DOCKER_REPO }}@%s ')
docker buildx imagetools inspect ${{ vars.DOCKER_REPO }}:${{ steps.metadata.outputs.version }}
working-directory: ${{ runner.temp }}
# Trigger downstream release process
trigger:
runs-on: ubuntu-latest
environment: release
needs: [darwin-build, windows-build, windows-depends]
steps:
- name: Trigger downstream release process
run: |
curl -L \
-X POST \
-H "Accept: application/vnd.github+json" \
-H "Authorization: Bearer ${{ secrets.RELEASE_TOKEN }}" \
-H "X-GitHub-Api-Version: 2022-11-28" \
https://api.github.com/repos/ollama/${{ vars.RELEASE_REPO }}/dispatches \
-d "{\"event_type\": \"trigger-workflow\", \"client_payload\": {\"run_id\": \"${GITHUB_RUN_ID}\", \"version\": \"${GITHUB_REF_NAME#v}\"}}"
# Aggregate all the assets and ship a release
release:
needs: [darwin-sign, windows-sign, linux-build]
runs-on: linux
environment: release
needs: [darwin-build, windows-build, windows-depends, linux-build]
permissions:
contents: write
env:
GH_TOKEN: ${{ github.token }}
steps:
- uses: actions/checkout@v4
- uses: actions/download-artifact@v4
with:
name: dist-darwin
path: dist
- uses: actions/download-artifact@v4
with:
name: dist-windows
path: dist
- uses: actions/download-artifact@v4
with:
pattern: dist-linux-*
path: dist
merge-multiple: true
- run: find . -type f -not -name 'sha256sum.txt' | xargs sha256sum | tee sha256sum.txt
working-directory: dist
- name: Create or update Release
- name: Create or update Release for tag
run: |
RELEASE_VERSION="$(echo ${GITHUB_REF_NAME} | cut -f1 -d-)"
echo "Looking for existing release for ${RELEASE_VERSION}"
OLD_TAG=$(gh release ls --json name,tagName | jq -r ".[] | select(.name == \"${RELEASE_VERSION}\") | .tagName")
if [ -n "$OLD_TAG" ]; then
@@ -491,5 +372,12 @@ jobs:
--generate-notes \
--prerelease
fi
echo "Uploading artifacts for tag ${GITHUB_REF_NAME}"
gh release upload ${GITHUB_REF_NAME} dist/* --clobber
- name: Trigger downstream release process
run: |
curl -L \
-X POST \
-H "Accept: application/vnd.github+json" \
-H "Authorization: Bearer ${{ secrets.RELEASE_TOKEN }}" \
-H "X-GitHub-Api-Version: 2022-11-28" \
https://api.github.com/repos/ollama/${{ vars.RELEASE_REPO }}/dispatches \
-d "{\"event_type\": \"trigger-workflow\", \"client_payload\": {\"run_id\": \"${GITHUB_RUN_ID}\", \"version\": \"${GITHUB_REF_NAME#v}\", \"origin\": \"${GITHUB_REPOSITORY}\", \"publish\": \"1\"}}"

View File

@@ -36,7 +36,7 @@ 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/**') | tee -a $GITHUB_OUTPUT
echo changed=$(changed 'llama/llama.cpp/**/*' 'ml/backend/ggml/ggml/**/*') | tee -a $GITHUB_OUTPUT
linux:
needs: [changes]
@@ -46,7 +46,7 @@ jobs:
include:
- preset: CPU
- preset: CUDA
container: nvidia/cuda:11.8.0-devel-ubuntu22.04
container: nvidia/cuda:12.8.1-devel-ubuntu22.04
flags: '-DCMAKE_CUDA_ARCHITECTURES=87'
- preset: ROCm
container: rocm/dev-ubuntu-22.04:6.1.2
@@ -78,11 +78,11 @@ jobs:
include:
- preset: CPU
- preset: CUDA
install: https://developer.download.nvidia.com/compute/cuda/11.3.1/local_installers/cuda_11.3.1_465.89_win10.exe
install: https://developer.download.nvidia.com/compute/cuda/12.8.0/local_installers/cuda_12.8.0_571.96_windows.exe
flags: '-DCMAKE_CUDA_ARCHITECTURES=80'
- 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'
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"'
runs-on: windows
steps:
- run: |
@@ -102,7 +102,7 @@ jobs:
$ErrorActionPreference = "Stop"
if ("${{ steps.cache-install.outputs.cache-hit }}" -ne 'true') {
Invoke-WebRequest -Uri "${{ matrix.install }}" -OutFile "install.exe"
Start-Process -FilePath .\install.exe -ArgumentList (@("-s", "cudart_11.3", "nvcc_11.3", "cublas_11.3", "cublas_dev_11.3")) -NoNewWindow -Wait
Start-Process -FilePath .\install.exe -ArgumentList (@("-s", "cudart_12.8", "nvcc_12.8", "cublas_12.8", "cublas_dev_12.8")) -NoNewWindow -Wait
}
$cudaPath = (Resolve-Path "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\*").path
@@ -120,6 +120,9 @@ jobs:
echo "$hipPath\bin" | Out-File -FilePath $env:GITHUB_PATH -Encoding utf8 -Append
echo "CC=$hipPath\bin\clang.exe" | Out-File -FilePath $env:GITHUB_ENV -Append
echo "CXX=$hipPath\bin\clang++.exe" | Out-File -FilePath $env:GITHUB_ENV -Append
echo "HIPCXX=$hipPath\bin\clang++.exe" | Out-File -FilePath $env:GITHUB_ENV -Append
echo "HIP_PLATFORM=amd" | Out-File -FilePath $env:GITHUB_ENV -Append
echo "CMAKE_PREFIX_PATH=$hipPath" | Out-File -FilePath $env:GITHUB_ENV -Append
- if: ${{ !cancelled() && steps.cache-install.outputs.cache-hit != 'true' }}
uses: actions/cache/save@v4
with:
@@ -133,8 +136,8 @@ jobs:
path: ${{ github.workspace }}\.ccache
key: ccache-${{ runner.os }}-${{ runner.arch }}-${{ matrix.preset }}
- run: |
Import-Module 'C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\Common7\Tools\Microsoft.VisualStudio.DevShell.dll'
Enter-VsDevShell -VsInstallPath 'C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise' -SkipAutomaticLocation -DevCmdArguments '-arch=x64 -no_logo'
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 }}" ${{ matrix.flags }}
cmake --build --parallel --preset "${{ matrix.preset }}"
env: