llm: consider head_dim in llama arch (#5817)

Adjust windows ROCm discovery

.dockerignore

@@ -1,8 +1,9 @@
 .vscode
 ollama
 app
+macapp
 dist
 llm/llama.cpp
 .env
 .cache
-test_data
+test_data

.gitattributes

@@ -0,0 +1 @@
+llm/ext_server/* linguist-vendored

.github/ISSUE_TEMPLATE/10_bug_report.yml

@@ -0,0 +1,60 @@
+name: Bug report
+labels: [bug]
+description: Something isn't working right.
+body:
+  - type: textarea
+    id: description
+    attributes:
+      label: What is the issue?
+      description: What happened? What did you expect to happen?
+    validations:
+      required: true
+  - type: dropdown
+    id: os
+    attributes:
+      label: OS
+      description: Which operating system are you using?
+      multiple: true
+      options:
+        - Linux
+        - macOS
+        - Windows
+        - Docker
+        - WSL2
+    validations:
+      required: false
+  - type: dropdown
+    id: gpu
+    attributes:
+      label: GPU
+      description: Which GPU are you using?
+      multiple: true
+      options:
+        - Nvidia
+        - AMD
+        - Intel
+        - Apple
+        - Other
+    validations:
+      required: false
+  - type: dropdown
+    id: cpu
+    attributes:
+      label: CPU
+      description: Which CPU are you using?
+      multiple: true
+      options:
+        - Intel
+        - AMD
+        - Apple
+        - Other
+    validations:
+      required: false
+  - type: input
+    id: version
+    attributes:
+      label: Ollama version
+      description: What version of Ollama are you using? (`ollama --version`)
+      placeholder: e.g., 0.1.32
+    validations:
+      required: false

.github/ISSUE_TEMPLATE/20_feature_request.md

@@ -0,0 +1,6 @@
+---
+name: Feature request
+about: Request a new feature
+labels: feature request
+---
+

.github/ISSUE_TEMPLATE/30_model_request.md

@@ -0,0 +1,5 @@
+---
+name: Model request
+about: Request support for a new model to be added to Ollama
+labels: model request
+---

.github/ISSUE_TEMPLATE/config.yml

@@ -0,0 +1,8 @@
+blank_issues_enabled: true
+contact_links:
+  - name: Help
+    url: https://discord.com/invite/ollama
+    about: Please join our Discord server for help using Ollama
+  - name: Troubleshooting
+    url: https://github.com/ollama/ollama/blob/main/docs/faq.md#faq
+    about: See the FAQ for common issues and solutions

.github/workflows/latest.yaml

@@ -0,0 +1,24 @@
+name: latest
+
+on:
+  release:
+    types: [released]
+
+jobs:
+  update-latest:
+    environment: release
+    runs-on: linux
+    steps:
+      - uses: actions/checkout@v4
+      - name: Login to Docker Hub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ vars.DOCKER_USER }}
+          password: ${{ secrets.DOCKER_ACCESS_TOKEN }}
+      - name: Tag images as latest
+        env:
+          PUSH: "1"
+        shell: bash
+        run: |
+          export "VERSION=${GITHUB_REF_NAME#v}"
+          ./scripts/tag_latest.sh

.github/workflows/release.yaml

@@ -0,0 +1,480 @@
+name: release
+
+on:
+  push:
+    tags:
+      - 'v*'
+
+jobs:
+  # Full build of the Mac assets
+  build-darwin:
+    runs-on: macos-12
+    environment: release
+    steps:
+      - uses: actions/checkout@v4
+      - name: Set Version
+        shell: bash
+        run: |
+          echo "VERSION=${GITHUB_REF_NAME#v}" >> $GITHUB_ENV
+          echo "RELEASE_VERSION=$(echo ${GITHUB_REF_NAME} | cut -f1 -d-)" >> $GITHUB_ENV
+      - name: key
+        env:
+          MACOS_SIGNING_KEY: ${{ secrets.MACOS_SIGNING_KEY }}
+          MACOS_SIGNING_KEY_PASSWORD: ${{ secrets.MACOS_SIGNING_KEY_PASSWORD }}
+        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
+      - uses: actions/setup-go@v5
+        with:
+          go-version-file: go.mod
+          cache: true
+      - name: Build Darwin
+        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_13.4.1.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk
+          DEVELOPER_DIR: /Applications/Xcode_13.4.1.app/Contents/Developer
+        run: |
+          ./scripts/build_darwin.sh
+
+      - uses: actions/upload-artifact@v4
+        with:
+          name: dist-darwin
+          path: |
+            dist/*arwin*
+            !dist/*-cov
+
+  # Windows builds take a long time to both install the dependencies and build, so parallelize
+  # CPU generation step
+  generate-windows-cpu:
+    environment: release
+    runs-on: windows
+    env:
+      KEY_CONTAINER: ${{ vars.KEY_CONTAINER }}
+    steps:
+      - uses: actions/checkout@v4
+      - name: Set Version
+        shell: bash
+        run: echo "VERSION=${GITHUB_REF_NAME#v}" >> $GITHUB_ENV
+      - uses: 'google-github-actions/auth@v2'
+        with:
+          project_id: 'ollama'
+          credentials_json: '${{ secrets.GOOGLE_SIGNING_CREDENTIALS }}'
+      - run: echo "${{ vars.OLLAMA_CERT }}" > ollama_inc.crt
+      - name: install Windows SDK 8.1 to get signtool
+        run: |
+          $ErrorActionPreference = "Stop"
+          write-host "downloading SDK"
+          Invoke-WebRequest -Uri "https://go.microsoft.com/fwlink/p/?LinkId=323507" -OutFile "${env:RUNNER_TEMP}\sdksetup.exe"
+          Start-Process "${env:RUNNER_TEMP}\sdksetup.exe" -ArgumentList @("/q") -NoNewWindow -Wait
+          write-host "Win SDK 8.1 installed"
+          gci -path 'C:\Program Files (x86)\Windows Kits\' -r -fi 'signtool.exe'
+      - name: install signing plugin
+        run: |
+          $ErrorActionPreference = "Stop"
+          write-host "downloading plugin"
+          Invoke-WebRequest -Uri "https://github.com/GoogleCloudPlatform/kms-integrations/releases/download/cng-v1.0/kmscng-1.0-windows-amd64.zip" -OutFile "${env:RUNNER_TEMP}\plugin.zip"
+          Expand-Archive -Path "${env:RUNNER_TEMP}\plugin.zip" -DestinationPath ${env:RUNNER_TEMP}\plugin\
+          write-host "Installing plugin"
+          & "${env:RUNNER_TEMP}\plugin\*\kmscng.msi" /quiet
+          write-host "plugin installed"
+      - uses: actions/setup-go@v5
+        with:
+          go-version-file: go.mod
+          cache: true
+      - run: go get ./...
+      - run: |
+          $gopath=(get-command go).source | split-path -parent
+          & "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\Common7\Tools\Launch-VsDevShell.ps1"
+          cd $env:GITHUB_WORKSPACE
+          $env:CMAKE_SYSTEM_VERSION="10.0.22621.0"
+          $env:PATH="$gopath;$env:PATH"
+          go generate -x ./...
+        name: go generate
+      - uses: actions/upload-artifact@v4
+        with:
+          name: generate-windows-cpu
+          path: |
+            llm/build/**/bin/*
+            llm/build/**/*.a
+            dist/windows-amd64/**
+
+  # ROCm generation step
+  generate-windows-rocm:
+    environment: release
+    runs-on: windows
+    env:
+      KEY_CONTAINER: ${{ vars.KEY_CONTAINER }}
+    steps:
+      - uses: actions/checkout@v4
+      - name: Set Version
+        shell: bash
+        run: echo "VERSION=${GITHUB_REF_NAME#v}" >> $GITHUB_ENV
+      - uses: 'google-github-actions/auth@v2'
+        with:
+          project_id: 'ollama'
+          credentials_json: '${{ secrets.GOOGLE_SIGNING_CREDENTIALS }}'
+      - run: echo "${{ vars.OLLAMA_CERT }}" > ollama_inc.crt
+      - name: install Windows SDK 8.1 to get signtool
+        run: |
+          $ErrorActionPreference = "Stop"
+          write-host "downloading SDK"
+          Invoke-WebRequest -Uri "https://go.microsoft.com/fwlink/p/?LinkId=323507" -OutFile "${env:RUNNER_TEMP}\sdksetup.exe"
+          Start-Process "${env:RUNNER_TEMP}\sdksetup.exe" -ArgumentList @("/q") -NoNewWindow -Wait
+          write-host "Win SDK 8.1 installed"
+          gci -path 'C:\Program Files (x86)\Windows Kits\' -r -fi 'signtool.exe'
+      - name: install signing plugin
+        run: |
+          $ErrorActionPreference = "Stop"
+          write-host "downloading plugin"
+          Invoke-WebRequest -Uri "https://github.com/GoogleCloudPlatform/kms-integrations/releases/download/cng-v1.0/kmscng-1.0-windows-amd64.zip" -OutFile "${env:RUNNER_TEMP}\plugin.zip"
+          Expand-Archive -Path "${env:RUNNER_TEMP}\plugin.zip" -DestinationPath ${env:RUNNER_TEMP}\plugin\
+          write-host "Installing plugin"
+          & "${env:RUNNER_TEMP}\plugin\*\kmscng.msi" /quiet
+          write-host "plugin installed"
+      - uses: actions/setup-go@v5
+        with:
+          go-version-file: go.mod
+          cache: true
+      - name: 'Install ROCm'
+        run: |
+          $ErrorActionPreference = "Stop"
+          write-host "downloading AMD HIP Installer"
+          Invoke-WebRequest -Uri "https://download.amd.com/developer/eula/rocm-hub/AMD-Software-PRO-Edition-24.Q3-WinSvr2022-For-HIP.exe" -OutFile "${env:RUNNER_TEMP}\rocm-install.exe"
+          write-host "Installing AMD HIP"
+          Start-Process "${env:RUNNER_TEMP}\rocm-install.exe" -ArgumentList '-install' -NoNewWindow -Wait
+          write-host "Completed AMD HIP"
+      - name: 'Verify ROCm'
+        run: |
+          & 'C:\Program Files\AMD\ROCm\*\bin\clang.exe' --version
+      - run: go get ./...
+      - run: |
+          $gopath=(get-command go).source | split-path -parent
+          & "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\Common7\Tools\Launch-VsDevShell.ps1"
+          cd $env:GITHUB_WORKSPACE
+          $env:CMAKE_SYSTEM_VERSION="10.0.22621.0"
+          $env:PATH="$gopath;$env:PATH"
+          $env:OLLAMA_SKIP_CPU_GENERATE="1"
+          $env:HIP_PATH=$(Resolve-Path 'C:\Program Files\AMD\ROCm\*\bin\clang.exe' | split-path | split-path)
+          go generate -x ./...
+        name: go generate
+      - name: 'gather rocm dependencies'
+        run: |
+          $HIP_PATH=$(Resolve-Path 'C:\Program Files\AMD\ROCm\*\bin\clang.exe' | split-path | split-path)
+          md "dist\deps\bin\rocblas\library"
+          cp "${HIP_PATH}\bin\hipblas.dll" "dist\deps\bin\"
+          cp "${HIP_PATH}\bin\rocblas.dll" "dist\deps\bin\"
+          cp "${HIP_PATH}\bin\rocblas\library\*" "dist\deps\bin\rocblas\library\"
+      - uses: actions/upload-artifact@v4
+        with:
+          name: generate-windows-rocm
+          path: |
+            llm/build/**/bin/*
+            dist/windows-amd64/**
+      - uses: actions/upload-artifact@v4
+        with:
+          name: windows-rocm-deps
+          path: dist/deps/*
+
+  # CUDA generation step
+  generate-windows-cuda:
+    environment: release
+    runs-on: windows
+    env:
+      KEY_CONTAINER: ${{ vars.KEY_CONTAINER }}
+    steps:
+      - uses: actions/checkout@v4
+      - name: Set Version
+        shell: bash
+        run: echo "VERSION=${GITHUB_REF_NAME#v}" >> $GITHUB_ENV
+      - uses: 'google-github-actions/auth@v2'
+        with:
+          project_id: 'ollama'
+          credentials_json: '${{ secrets.GOOGLE_SIGNING_CREDENTIALS }}'
+      - run: echo "${{ vars.OLLAMA_CERT }}" > ollama_inc.crt
+      - name: install Windows SDK 8.1 to get signtool
+        run: |
+          $ErrorActionPreference = "Stop"
+          write-host "downloading SDK"
+          Invoke-WebRequest -Uri "https://go.microsoft.com/fwlink/p/?LinkId=323507" -OutFile "${env:RUNNER_TEMP}\sdksetup.exe"
+          Start-Process "${env:RUNNER_TEMP}\sdksetup.exe" -ArgumentList @("/q") -NoNewWindow -Wait
+          write-host "Win SDK 8.1 installed"
+          gci -path 'C:\Program Files (x86)\Windows Kits\' -r -fi 'signtool.exe'
+      - name: install signing plugin
+        run: |
+          $ErrorActionPreference = "Stop"
+          write-host "downloading plugin"
+          Invoke-WebRequest -Uri "https://github.com/GoogleCloudPlatform/kms-integrations/releases/download/cng-v1.0/kmscng-1.0-windows-amd64.zip" -OutFile "${env:RUNNER_TEMP}\plugin.zip"
+          Expand-Archive -Path "${env:RUNNER_TEMP}\plugin.zip" -DestinationPath ${env:RUNNER_TEMP}\plugin\
+          write-host "Installing plugin"
+          & "${env:RUNNER_TEMP}\plugin\*\kmscng.msi" /quiet
+          write-host "plugin installed"
+      - uses: actions/setup-go@v5
+        with:
+          go-version-file: go.mod
+          cache: true
+      - name: 'Install CUDA'
+        run: |
+          $ErrorActionPreference = "Stop"
+          write-host "downloading CUDA Installer"
+          Invoke-WebRequest -Uri "https://developer.download.nvidia.com/compute/cuda/11.3.1/local_installers/cuda_11.3.1_465.89_win10.exe" -OutFile "${env:RUNNER_TEMP}\cuda-install.exe"
+          write-host "Installing CUDA"
+          Start-Process "${env:RUNNER_TEMP}\cuda-install.exe" -ArgumentList '-s' -NoNewWindow -Wait
+          write-host "Completed CUDA"
+          $cudaPath=((resolve-path "c:\Program Files\NVIDIA*\CUDA\v*\bin\nvcc.exe")[0].path | split-path | split-path)
+          $cudaVer=($cudaPath | split-path -leaf ) -replace 'v(\d+).(\d+)', '$1_$2' 
+          echo "$cudaPath\bin" >> $env:GITHUB_PATH
+          echo "CUDA_PATH=$cudaPath" >> $env:GITHUB_ENV
+          echo "CUDA_PATH_V${cudaVer}=$cudaPath" >> $env:GITHUB_ENV
+          echo "CUDA_PATH_VX_Y=CUDA_PATH_V${cudaVer}" >> $env:GITHUB_ENV
+      - name: 'Verify CUDA'
+        run: nvcc -V
+      - run: go get ./...
+      - name: go generate
+        run: |
+          $gopath=(get-command go).source | split-path -parent
+          $cudabin=(get-command nvcc).source | split-path
+          & "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\Common7\Tools\Launch-VsDevShell.ps1"
+          cd $env:GITHUB_WORKSPACE
+          $env:CMAKE_SYSTEM_VERSION="10.0.22621.0"
+          $env:PATH="$gopath;$cudabin;$env:PATH"
+          $env:OLLAMA_SKIP_CPU_GENERATE="1"
+          go generate -x ./...
+      - name: 'gather cuda dependencies'
+        run: |
+          $NVIDIA_DIR=(resolve-path 'C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\*\bin\')[0]
+          md "dist\deps"
+          cp "${NVIDIA_DIR}\cudart64_*.dll" "dist\deps\"
+          cp "${NVIDIA_DIR}\cublas64_*.dll" "dist\deps\"
+          cp "${NVIDIA_DIR}\cublasLt64_*.dll" "dist\deps\"
+      - uses: actions/upload-artifact@v4
+        with:
+          name: generate-windows-cuda
+          path: |
+            llm/build/**/bin/*
+            dist/windows-amd64/**
+      - uses: actions/upload-artifact@v4
+        with:
+          name: windows-cuda-deps
+          path: dist/deps/*
+
+  # Import the prior generation steps and build the final windows assets
+  build-windows:
+    environment: release
+    runs-on: windows
+    needs:
+      - generate-windows-cuda
+      - generate-windows-rocm
+      - generate-windows-cpu
+    env:
+      KEY_CONTAINER: ${{ vars.KEY_CONTAINER }}
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: recursive
+      - name: Set Version
+        shell: bash
+        run: echo "VERSION=${GITHUB_REF_NAME#v}" >> $GITHUB_ENV
+      - uses: 'google-github-actions/auth@v2'
+        with:
+          project_id: 'ollama'
+          credentials_json: '${{ secrets.GOOGLE_SIGNING_CREDENTIALS }}'
+      - run: echo "${{ vars.OLLAMA_CERT }}" > ollama_inc.crt
+      - name: install Windows SDK 8.1 to get signtool
+        run: |
+          $ErrorActionPreference = "Stop"
+          write-host "downloading SDK"
+          Invoke-WebRequest -Uri "https://go.microsoft.com/fwlink/p/?LinkId=323507" -OutFile "${env:RUNNER_TEMP}\sdksetup.exe"
+          Start-Process "${env:RUNNER_TEMP}\sdksetup.exe" -ArgumentList @("/q") -NoNewWindow -Wait
+          write-host "Win SDK 8.1 installed"
+          gci -path 'C:\Program Files (x86)\Windows Kits\' -r -fi 'signtool.exe'
+      - name: install signing plugin
+        run: |
+          $ErrorActionPreference = "Stop"
+          write-host "downloading plugin"
+          Invoke-WebRequest -Uri "https://github.com/GoogleCloudPlatform/kms-integrations/releases/download/cng-v1.0/kmscng-1.0-windows-amd64.zip" -OutFile "${env:RUNNER_TEMP}\plugin.zip"
+          Expand-Archive -Path "${env:RUNNER_TEMP}\plugin.zip" -DestinationPath ${env:RUNNER_TEMP}\plugin\
+          write-host "Installing plugin"
+          & "${env:RUNNER_TEMP}\plugin\*\kmscng.msi" /quiet
+          write-host "plugin installed"
+      - uses: actions/setup-go@v5
+        with:
+          go-version-file: go.mod
+          cache: true
+      - run: go get
+      - uses: actions/download-artifact@v4
+        with:
+          name: generate-windows-cpu
+      - uses: actions/download-artifact@v4
+        with:
+          name: generate-windows-cuda
+      - uses: actions/download-artifact@v4
+        with:
+          name: windows-cuda-deps
+      - uses: actions/download-artifact@v4
+        with:
+          name: windows-rocm-deps
+      - uses: actions/download-artifact@v4
+        with:
+          name: generate-windows-rocm
+      - run: dir llm/build
+      - run: |
+          $gopath=(get-command go).source | split-path -parent
+          & "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\Common7\Tools\Launch-VsDevShell.ps1"
+          cd $env:GITHUB_WORKSPACE
+          $env:CMAKE_SYSTEM_VERSION="10.0.22621.0"
+          $env:PATH="$gopath;$env:PATH"
+          $env:OLLAMA_SKIP_GENERATE="1"
+          & .\scripts\build_windows.ps1
+      - uses: actions/upload-artifact@v4
+        with:
+          name: dist-windows
+          path: |
+            dist/OllamaSetup.exe
+            dist/ollama-windows-*.zip
+
+  # Linux x86 assets built using the container based build
+  build-linux-amd64:
+    environment: release
+    runs-on: linux
+    env:
+      OLLAMA_SKIP_MANIFEST_CREATE: '1'
+      BUILD_ARCH: amd64
+      PUSH: '1'
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: recursive
+      - name: Set Version
+        shell: bash
+        run: echo "VERSION=${GITHUB_REF_NAME#v}" >> $GITHUB_ENV
+      - name: Login to Docker Hub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ vars.DOCKER_USER }}
+          password: ${{ secrets.DOCKER_ACCESS_TOKEN }}
+      - run: |
+          ./scripts/build_linux.sh
+          ./scripts/build_docker.sh
+          mv dist/deps/* dist/
+      - uses: actions/upload-artifact@v4
+        with:
+          name: dist-linux-amd64
+          path: |
+            dist/*linux*
+            !dist/*-cov
+
+  # Linux ARM assets built using the container based build
+  # (at present, docker isn't pre-installed on arm ubunutu images)
+  build-linux-arm64:
+    environment: release
+    runs-on: linux-arm64
+    env:
+      OLLAMA_SKIP_MANIFEST_CREATE: '1'
+      BUILD_ARCH: arm64
+      PUSH: '1'
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          submodules: recursive
+      - name: Set Version
+        shell: bash
+        run: echo "VERSION=${GITHUB_REF_NAME#v}" >> $GITHUB_ENV
+      - name: 'Install Docker'
+        run: |
+          # Add Docker's official GPG key:
+          env
+          uname -a
+          sudo apt-get update
+          sudo apt-get install -y ca-certificates curl
+          sudo install -m 0755 -d /etc/apt/keyrings
+          sudo curl -fsSL https://download.docker.com/linux/ubuntu/gpg -o /etc/apt/keyrings/docker.asc
+          sudo chmod a+r /etc/apt/keyrings/docker.asc
+
+          # Add the repository to Apt sources:
+          echo \
+            "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.asc] https://download.docker.com/linux/ubuntu \
+            $(. /etc/os-release && echo "$VERSION_CODENAME") stable" | \
+            sudo tee /etc/apt/sources.list.d/docker.list > /dev/null
+          sudo apt-get update
+          sudo apt-get install -y docker-ce docker-ce-cli containerd.io
+          sudo usermod -aG docker $USER
+          sudo apt-get install acl
+          sudo setfacl --modify user:$USER:rw /var/run/docker.sock
+      - name: Login to Docker Hub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ vars.DOCKER_USER }}
+          password: ${{ secrets.DOCKER_ACCESS_TOKEN }}
+      - run: |
+          ./scripts/build_linux.sh
+          ./scripts/build_docker.sh
+      - uses: actions/upload-artifact@v4
+        with:
+          name: dist-linux-arm64
+          path: |
+            dist/*linux*
+            !dist/*-cov
+
+  # Aggregate all the assets and ship a release
+  release:
+    needs:
+      - build-darwin
+      - build-windows
+      - build-linux-amd64
+      - build-linux-arm64
+    runs-on: linux
+    environment: release
+    permissions:
+      contents: write
+    env:
+      OLLAMA_SKIP_IMAGE_BUILD: '1'
+      PUSH: '1'
+      GH_TOKEN: ${{ github.token }}
+    steps:
+      - uses: actions/checkout@v4
+      - name: Set Version
+        shell: bash
+        run: |
+          echo "VERSION=${GITHUB_REF_NAME#v}" >> $GITHUB_ENV
+          echo "RELEASE_VERSION=$(echo ${GITHUB_REF_NAME} | cut -f1 -d-)" >> $GITHUB_ENV
+      - name: Login to Docker Hub
+        uses: docker/login-action@v3
+        with:
+          username: ${{ vars.DOCKER_USER }}
+          password: ${{ secrets.DOCKER_ACCESS_TOKEN }}
+      - run: ./scripts/build_docker.sh
+      - name: Retrieve built artifact
+        uses: actions/download-artifact@v4
+        with:
+          path: dist
+          pattern: dist-*
+          merge-multiple: true
+      - run: |
+          ls -lh dist/
+          (cd dist; sha256sum * > sha256sum.txt)
+          cat dist/sha256sum.txt
+      - name: Create or update Release
+        run: |
+          echo "Looking for existing release for ${{ env.RELEASE_VERSION }}"
+          OLD_TAG=$(gh release ls --json name,tagName | jq -r ".[] | select(.name == \"${{ env.RELEASE_VERSION }}\") | .tagName")
+          if [ -n "$OLD_TAG" ]; then
+            echo "Updating release ${{ env.RELEASE_VERSION }} to point to new tag ${GITHUB_REF_NAME}"
+            gh release edit ${OLD_TAG} --tag ${GITHUB_REF_NAME}
+          else
+            echo "Creating new release ${{ env.RELEASE_VERSION }} pointing to tag ${GITHUB_REF_NAME}"
+            gh release create ${GITHUB_REF_NAME} \
+              --title ${{ env.RELEASE_VERSION }} \
+              --draft \
+              --generate-notes \
+              --prerelease
+          fi
+          echo "Uploading artifacts for tag ${GITHUB_REF_NAME}"
+          gh release upload ${GITHUB_REF_NAME} dist/* --clobber

.github/workflows/test.yaml

@@ -1,35 +1,95 @@
 name: test
 
+concurrency:
+  # For PRs, later CI runs preempt previous ones. e.g. a force push on a PR
+  # cancels running CI jobs and starts all new ones.
+  #
+  # For non-PR pushes, concurrency.group needs to be unique for every distinct
+  # CI run we want to have happen. Use run_id, which in practice means all
+  # non-PR CI runs will be allowed to run without preempting each other.
+  group: ${{ github.workflow }}-$${{ github.pull_request.number || github.run_id }}
+  cancel-in-progress: true
+
 on:
   pull_request:
+    paths:
+      - '**/*'
+      - '!docs/**'
+      - '!README.md'
 
 jobs:
+  changes:
+    runs-on: ubuntu-latest
+    outputs:
+      GENERATE: ${{ steps.changes.outputs.GENERATE }}
+      GENERATE_CUDA: ${{ steps.changes.outputs.GENERATE_CUDA }}
+      GENERATE_ROCM: ${{ steps.changes.outputs.GENERATE_ROCM }}
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+      - id: changes
+        run: |
+          changed() {
+            git diff-tree -r --no-commit-id --name-only \
+              $(git merge-base ${{ github.event.pull_request.base.sha }} ${{ github.event.pull_request.head.sha }}) \
+              ${{ github.event.pull_request.head.sha }} \
+              | xargs python3 -c "import sys; from pathlib import Path; print(any(Path(x).match(glob) for x in sys.argv[1:] for glob in '$*'.split(' ')))"
+          }
+
+          {
+            echo GENERATE=$(changed 'llm/llama.cpp' 'llm/patches/**' 'llm/ext_server/**' 'llm/generate/**')
+            echo GENERATE_CUDA=$(changed 'llm/llama.cpp' 'llm/patches/**' 'llm/ext_server/**' 'llm/generate/**')
+            echo GENERATE_ROCM=$(changed 'llm/llama.cpp' 'llm/patches/**' 'llm/ext_server/**' 'llm/generate/**')
+          } >>$GITHUB_OUTPUT
+
   generate:
+    needs: [changes]
+    if: ${{ needs.changes.outputs.GENERATE == 'True' }}
     strategy:
       matrix:
-        os: [ubuntu-latest, macos-latest, windows-latest]
+        os: [ubuntu-latest, macos-latest, windows-2019]
         arch: [amd64, arm64]
         exclude:
           - os: ubuntu-latest
             arch: arm64
-          - os: windows-latest
+          - os: windows-2019
             arch: arm64
     runs-on: ${{ matrix.os }}
     env:
       GOARCH: ${{ matrix.arch }}
+      CGO_ENABLED: '1'
     steps:
       - uses: actions/checkout@v4
       - uses: actions/setup-go@v5
         with:
-          go-version: '1.21'
+          go-version-file: go.mod
           cache: true
       - run: go get ./...
+      - run: |
+          $gopath=(get-command go).source | split-path -parent
+          $gccpath=(get-command gcc).source | split-path -parent
+          & "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\Common7\Tools\Launch-VsDevShell.ps1"
+          cd $env:GITHUB_WORKSPACE
+          $env:CMAKE_SYSTEM_VERSION="10.0.22621.0"
+          $env:PATH="$gopath;$gccpath;$env:PATH"
+          echo $env:PATH
+          go generate -x ./...
+        if: ${{ startsWith(matrix.os, 'windows-') }}
+        name: 'Windows Go Generate'
       - run: go generate -x ./...
+        if: ${{ ! startsWith(matrix.os, 'windows-') }}
+        name: 'Unix Go Generate'
+      - run: go build .
       - uses: actions/upload-artifact@v4
         with:
           name: ${{ matrix.os }}-${{ matrix.arch }}-libraries
-          path: llm/llama.cpp/build/**/lib/*
+          path: |
+            llm/build/**/bin/*
+            llm/build/**/*.a
   generate-cuda:
+    needs: [changes]
+    if: ${{ needs.changes.outputs.GENERATE_CUDA == 'True' }}
     strategy:
       matrix:
         cuda-version:
@@ -46,7 +106,7 @@ jobs:
       - uses: actions/checkout@v4
       - uses: actions/setup-go@v4
         with:
-          go-version: '1.21'
+          go-version-file: go.mod
           cache: true
       - run: go get ./...
       - run: |
@@ -57,13 +117,16 @@ jobs:
       - uses: actions/upload-artifact@v4
         with:
           name: cuda-${{ matrix.cuda-version }}-libraries
-          path: llm/llama.cpp/build/**/lib/*
+          path: |
+            llm/build/**/bin/*
+            dist/windows-amd64/**
   generate-rocm:
+    needs: [changes]
+    if: ${{ needs.changes.outputs.GENERATE_ROCM == 'True' }}
     strategy:
       matrix:
         rocm-version:
-          - '5.7.1'
-          - '6.0'
+          - '6.1.2'
     runs-on: linux
     container: rocm/dev-ubuntu-20.04:${{ matrix.rocm-version }}
     steps:
@@ -76,7 +139,7 @@ jobs:
       - uses: actions/checkout@v4
       - uses: actions/setup-go@v4
         with:
-          go-version: '1.21'
+          go-version-file: go.mod
           cache: true
       - run: go get ./...
       - run: |
@@ -87,73 +150,171 @@ jobs:
       - uses: actions/upload-artifact@v4
         with:
           name: rocm-${{ matrix.rocm-version }}-libraries
-          path: llm/llama.cpp/build/**/lib/*
+          path: |
+            llm/build/**/bin/*
+            dist/windows-amd64/**
+
+  # ROCm generation step
+  generate-windows-rocm:
+    needs: [changes]
+    if: ${{ needs.changes.outputs.GENERATE_ROCM == 'True' }}
+    runs-on: windows
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-go@v5
+        with:
+          go-version-file: go.mod
+          cache: true
+      - name: 'Install ROCm'
+        run: |
+          $ErrorActionPreference = "Stop"
+          write-host "downloading AMD HIP Installer"
+          Invoke-WebRequest -Uri "https://download.amd.com/developer/eula/rocm-hub/AMD-Software-PRO-Edition-24.Q3-WinSvr2022-For-HIP.exe" -OutFile "${env:RUNNER_TEMP}\rocm-install.exe"
+          write-host "Installing AMD HIP"
+          Start-Process "${env:RUNNER_TEMP}\rocm-install.exe" -ArgumentList '-install' -NoNewWindow -Wait
+          write-host "Completed AMD HIP"
+      - name: 'Verify ROCm'
+        run: |
+          & 'C:\Program Files\AMD\ROCm\*\bin\clang.exe' --version
+      - run: go get ./...
+      - run: |
+          $gopath=(get-command go).source | split-path -parent
+          & "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\Common7\Tools\Launch-VsDevShell.ps1"
+          cd $env:GITHUB_WORKSPACE
+          $env:CMAKE_SYSTEM_VERSION="10.0.22621.0"
+          $env:PATH="$gopath;$env:PATH"
+          $env:OLLAMA_SKIP_CPU_GENERATE="1"
+          $env:HIP_PATH=$(Resolve-Path 'C:\Program Files\AMD\ROCm\*\bin\clang.exe' | split-path | split-path)
+          go generate -x ./...
+        name: go generate
+        env:
+          OLLAMA_SKIP_CPU_GENERATE: '1'
+      # TODO - do we need any artifacts?
+
+  # CUDA generation step
+  generate-windows-cuda:
+    needs: [changes]
+    if: ${{ needs.changes.outputs.GENERATE_CUDA == 'True' }}
+    runs-on: windows
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-go@v5
+        with:
+          go-version-file: go.mod
+          cache: true
+      - name: 'Install CUDA'
+        run: |
+          $ErrorActionPreference = "Stop"
+          write-host "downloading CUDA Installer"
+          Invoke-WebRequest -Uri "https://developer.download.nvidia.com/compute/cuda/11.3.1/local_installers/cuda_11.3.1_465.89_win10.exe" -OutFile "${env:RUNNER_TEMP}\cuda-install.exe"
+          write-host "Installing CUDA"
+          Start-Process "${env:RUNNER_TEMP}\cuda-install.exe" -ArgumentList '-s' -NoNewWindow -Wait
+          write-host "Completed CUDA"
+          $cudaPath=((resolve-path "c:\Program Files\NVIDIA*\CUDA\v*\bin\nvcc.exe")[0].path | split-path | split-path)
+          $cudaVer=($cudaPath | split-path -leaf ) -replace 'v(\d+).(\d+)', '$1_$2' 
+          echo "$cudaPath\bin" >> $env:GITHUB_PATH
+          echo "CUDA_PATH=$cudaPath" >> $env:GITHUB_ENV
+          echo "CUDA_PATH_V${cudaVer}=$cudaPath" >> $env:GITHUB_ENV
+          echo "CUDA_PATH_VX_Y=CUDA_PATH_V${cudaVer}" >> $env:GITHUB_ENV
+      - name: 'Verify CUDA'
+        run: nvcc -V
+      - run: go get ./...
+      - name: go generate
+        run: |
+          $gopath=(get-command go).source | split-path -parent
+          $cudabin=(get-command nvcc).source | split-path
+          & "C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\Common7\Tools\Launch-VsDevShell.ps1"
+          cd $env:GITHUB_WORKSPACE
+          $env:CMAKE_SYSTEM_VERSION="10.0.22621.0"
+          $env:PATH="$gopath;$cudabin;$env:PATH"
+          $env:OLLAMA_SKIP_CPU_GENERATE="1"
+          go generate -x ./...
+        env:
+          OLLAMA_SKIP_CPU_GENERATE: '1'
+      # TODO - do we need any artifacts?
+
   lint:
     strategy:
       matrix:
-        os: [ubuntu-latest, macos-latest, windows-latest]
+        os: [ubuntu-latest, macos-latest, windows-2019]
         arch: [amd64, arm64]
         exclude:
           - os: ubuntu-latest
             arch: arm64
-          - os: windows-latest
+          - os: windows-2019
             arch: arm64
           - os: macos-latest
             arch: amd64
     runs-on: ${{ matrix.os }}
     env:
       GOARCH: ${{ matrix.arch }}
-      CGO_ENABLED: "1"
+      CGO_ENABLED: '1'
     steps:
       - uses: actions/checkout@v4
         with:
           submodules: recursive
       - uses: actions/setup-go@v5
         with:
-          go-version: '1.21'
+          go-version-file: go.mod
           cache: false
       - run: |
-          mkdir -p llm/llama.cpp/build/linux/${{ matrix.arch }}/stub/lib/
-          touch llm/llama.cpp/build/linux/${{ matrix.arch }}/stub/lib/stub.so
+          case ${{ matrix.arch }} in
+            amd64) echo ARCH=x86_64 ;;
+            arm64) echo ARCH=arm64 ;;
+          esac >>$GITHUB_ENV
+        shell: bash
+      - run: |
+          mkdir -p llm/build/linux/$ARCH/stub/bin
+          touch llm/build/linux/$ARCH/stub/bin/ollama_llama_server
         if: ${{ startsWith(matrix.os, 'ubuntu-') }}
       - run: |
-          mkdir -p llm/llama.cpp/build/darwin/${{ matrix.arch }}/stub/lib/
-          touch llm/llama.cpp/build/darwin/${{ matrix.arch }}/stub/lib/stub.dylib
-          touch llm/llama.cpp/ggml-metal.metal
+          mkdir -p llm/build/darwin/$ARCH/stub/bin
+          touch llm/build/darwin/$ARCH/stub/bin/ollama_llama_server
         if: ${{ startsWith(matrix.os, 'macos-') }}
-      - run: |
-          mkdir -p llm/llama.cpp/build/windows/${{ matrix.arch }}/stub/lib/
-          touch llm/llama.cpp/build/windows/${{ matrix.arch }}/stub/lib/stub.dll
-        if: ${{ startsWith(matrix.os, 'windows-') }}
-      - uses: golangci/golangci-lint-action@v3
+      - uses: golangci/golangci-lint-action@v6
+        with:
+          args: --timeout 8m0s -v ${{ startsWith(matrix.os, 'windows-') && '' || '--disable gofmt --disable goimports' }}
   test:
-    needs: generate
     strategy:
       matrix:
-        os: [ubuntu-latest, macos-latest, windows-latest]
+        os: [ubuntu-latest, macos-latest, windows-2019]
         arch: [amd64]
         exclude:
           - os: ubuntu-latest
             arch: arm64
-          - os: windows-latest
+          - os: windows-2019
             arch: arm64
     runs-on: ${{ matrix.os }}
     env:
       GOARCH: ${{ matrix.arch }}
-      CGO_ENABLED: "1"
+      CGO_ENABLED: '1'
+      OLLAMA_CPU_TARGET: 'static'
+      OLLAMA_SKIP_CPU_GENERATE: '1'
+      OLLAMA_SKIP_METAL_GENERATE: '1'
     steps:
       - uses: actions/checkout@v4
         with:
           submodules: recursive
       - uses: actions/setup-go@v5
         with:
-          go-version: '1.21'
+          go-version-file: go.mod
           cache: true
-      - run: go get
-      - uses: actions/download-artifact@v4
-        with:
-          name: ${{ matrix.os }}-${{ matrix.arch }}-libraries
-          path: llm/llama.cpp/build
+      - run: |
+          case ${{ matrix.arch }} in
+            amd64) echo ARCH=x86_64 ;;
+            arm64) echo ARCH=arm64 ;;
+          esac >>$GITHUB_ENV
+        shell: bash
+      - run: |
+          mkdir -p llm/build/linux/$ARCH/stub/bin
+          touch llm/build/linux/$ARCH/stub/bin/ollama_llama_server
+        if: ${{ startsWith(matrix.os, 'ubuntu-') }}
+      - run: |
+          mkdir -p llm/build/darwin/$ARCH/stub/bin
+          touch llm/build/darwin/$ARCH/stub/bin/ollama_llama_server
+        if: ${{ startsWith(matrix.os, 'macos-') }}
+        shell: bash
+      - run: go generate ./...
       - run: go build
       - run: go test -v ./...
       - uses: actions/upload-artifact@v4

.gitignore

@@ -10,4 +10,6 @@ ggml-metal.metal
 *.exe
 .idea
 test_data
-*.crt
+*.crt
+llm/build
+__debug_bin*

.golangci.yaml

@@ -9,19 +9,26 @@ linters:
     - contextcheck
     - exportloopref
     - gocheckcompilerdirectives
-    # FIXME: for some reason this errors on windows
+    # conditionally enable this on linux/macos
     # - gofmt
     # - goimports
+    - intrange
     - misspell
     - nilerr
+    - nolintlint
+    - nosprintfhostport
+    - testifylint
+    - unconvert
     - unused
-linters-settings:
-  errcheck:
-    # exclude the following functions since we don't generally
-    # need to be concerned with the returned errors
-    exclude-functions:
-      - encoding/binary.Read
-      - (*os.File).Seek
-      - (*bufio.Writer).WriteString
-      - (*github.com/spf13/pflag.FlagSet).Set
-      - (*github.com/jmorganca/ollama/llm.readSeekOffset).Seek
+    - wastedassign
+    - whitespace
+    - usestdlibvars
+severity:
+  default-severity: error
+  rules:
+    - linters:
+        - gofmt
+        - goimports
+        - intrange
+        - usestdlibvars
+      severity: info

Dockerfile

@@ -1,6 +1,8 @@
-ARG GOLANG_VERSION=1.21.3
+ARG GOLANG_VERSION=1.22.1
 ARG CMAKE_VERSION=3.22.1
+# this CUDA_VERSION corresponds with the one specified in docs/gpu.md
 ARG CUDA_VERSION=11.3.1
+ARG ROCM_VERSION=6.1.2
 
 # Copy the minimal context we need to run the generate scripts
 FROM scratch AS llm-code
@@ -13,44 +15,40 @@ ARG CMAKE_VERSION
 COPY ./scripts/rh_linux_deps.sh /
 RUN CMAKE_VERSION=${CMAKE_VERSION} sh /rh_linux_deps.sh
 ENV PATH /opt/rh/devtoolset-10/root/usr/bin:$PATH
-COPY --from=llm-code / /go/src/github.com/jmorganca/ollama/
-WORKDIR /go/src/github.com/jmorganca/ollama/llm/generate
+COPY --from=llm-code / /go/src/github.com/ollama/ollama/
+WORKDIR /go/src/github.com/ollama/ollama/llm/generate
 ARG CGO_CFLAGS
-RUN OLLAMA_SKIP_CPU_GENERATE=1 sh gen_linux.sh
+RUN OLLAMA_SKIP_STATIC_GENERATE=1 OLLAMA_SKIP_CPU_GENERATE=1 sh gen_linux.sh
 
 FROM --platform=linux/arm64 nvidia/cuda:$CUDA_VERSION-devel-rockylinux8 AS cuda-build-arm64
 ARG CMAKE_VERSION
 COPY ./scripts/rh_linux_deps.sh /
 RUN CMAKE_VERSION=${CMAKE_VERSION} sh /rh_linux_deps.sh
 ENV PATH /opt/rh/gcc-toolset-10/root/usr/bin:$PATH
-COPY --from=llm-code / /go/src/github.com/jmorganca/ollama/
-WORKDIR /go/src/github.com/jmorganca/ollama/llm/generate
+COPY --from=llm-code / /go/src/github.com/ollama/ollama/
+WORKDIR /go/src/github.com/ollama/ollama/llm/generate
 ARG CGO_CFLAGS
-RUN OLLAMA_SKIP_CPU_GENERATE=1 sh gen_linux.sh
+RUN OLLAMA_SKIP_STATIC_GENERATE=1 OLLAMA_SKIP_CPU_GENERATE=1 sh gen_linux.sh
 
-FROM --platform=linux/amd64 rocm/dev-centos-7:5.7.1-complete AS rocm-5-build-amd64
+FROM --platform=linux/amd64 rocm/dev-centos-7:${ROCM_VERSION}-complete AS rocm-build-amd64
 ARG CMAKE_VERSION
 COPY ./scripts/rh_linux_deps.sh /
 RUN CMAKE_VERSION=${CMAKE_VERSION} sh /rh_linux_deps.sh
 ENV PATH /opt/rh/devtoolset-10/root/usr/bin:$PATH
 ENV LIBRARY_PATH /opt/amdgpu/lib64
-COPY --from=llm-code / /go/src/github.com/jmorganca/ollama/
-WORKDIR /go/src/github.com/jmorganca/ollama/llm/generate
+COPY --from=llm-code / /go/src/github.com/ollama/ollama/
+WORKDIR /go/src/github.com/ollama/ollama/llm/generate
 ARG CGO_CFLAGS
 ARG AMDGPU_TARGETS
-RUN OLLAMA_SKIP_CPU_GENERATE=1 sh gen_linux.sh
+RUN OLLAMA_SKIP_STATIC_GENERATE=1 OLLAMA_SKIP_CPU_GENERATE=1 sh gen_linux.sh
+RUN mkdir /tmp/scratch && \
+    for dep in $(zcat /go/src/github.com/ollama/ollama/llm/build/linux/x86_64/rocm*/bin/deps.txt.gz) ; do \
+        cp ${dep} /tmp/scratch/ || exit 1 ; \
+    done && \
+    (cd /opt/rocm/lib && tar cf - rocblas/library) | (cd /tmp/scratch/ && tar xf - ) && \
+    mkdir -p /go/src/github.com/ollama/ollama/dist/deps/ && \
+    (cd /tmp/scratch/ && tar czvf /go/src/github.com/ollama/ollama/dist/deps/ollama-linux-amd64-rocm.tgz . )
 
-FROM --platform=linux/amd64 rocm/dev-centos-7:6.0-complete AS rocm-6-build-amd64
-ARG CMAKE_VERSION
-COPY ./scripts/rh_linux_deps.sh /
-RUN CMAKE_VERSION=${CMAKE_VERSION} sh /rh_linux_deps.sh
-ENV PATH /opt/rh/devtoolset-10/root/usr/bin:$PATH
-ENV LIBRARY_PATH /opt/amdgpu/lib64
-COPY --from=llm-code / /go/src/github.com/jmorganca/ollama/
-WORKDIR /go/src/github.com/jmorganca/ollama/llm/generate
-ARG CGO_CFLAGS
-ARG AMDGPU_TARGETS
-RUN OLLAMA_SKIP_CPU_GENERATE=1 sh gen_linux.sh
 
 FROM --platform=linux/amd64 centos:7 AS cpu-builder-amd64
 ARG CMAKE_VERSION
@@ -58,68 +56,76 @@ ARG GOLANG_VERSION
 COPY ./scripts/rh_linux_deps.sh /
 RUN CMAKE_VERSION=${CMAKE_VERSION} GOLANG_VERSION=${GOLANG_VERSION} sh /rh_linux_deps.sh
 ENV PATH /opt/rh/devtoolset-10/root/usr/bin:$PATH
-COPY --from=llm-code / /go/src/github.com/jmorganca/ollama/
+COPY --from=llm-code / /go/src/github.com/ollama/ollama/
 ARG OLLAMA_CUSTOM_CPU_DEFS
 ARG CGO_CFLAGS
-WORKDIR /go/src/github.com/jmorganca/ollama/llm/generate
+WORKDIR /go/src/github.com/ollama/ollama/llm/generate
 
+FROM --platform=linux/amd64 cpu-builder-amd64 AS static-build-amd64
+RUN OLLAMA_CPU_TARGET="static" sh gen_linux.sh
 FROM --platform=linux/amd64 cpu-builder-amd64 AS cpu-build-amd64
-RUN OLLAMA_CPU_TARGET="cpu" sh gen_linux.sh
+RUN OLLAMA_SKIP_STATIC_GENERATE=1 OLLAMA_CPU_TARGET="cpu" sh gen_linux.sh
 FROM --platform=linux/amd64 cpu-builder-amd64 AS cpu_avx-build-amd64
-RUN OLLAMA_CPU_TARGET="cpu_avx" sh gen_linux.sh
+RUN OLLAMA_SKIP_STATIC_GENERATE=1 OLLAMA_CPU_TARGET="cpu_avx" sh gen_linux.sh
 FROM --platform=linux/amd64 cpu-builder-amd64 AS cpu_avx2-build-amd64
-RUN OLLAMA_CPU_TARGET="cpu_avx2" sh gen_linux.sh
+RUN OLLAMA_SKIP_STATIC_GENERATE=1 OLLAMA_CPU_TARGET="cpu_avx2" sh gen_linux.sh
 
-FROM --platform=linux/arm64 centos:7 AS cpu-build-arm64
+FROM --platform=linux/arm64 rockylinux:8 AS cpu-builder-arm64
 ARG CMAKE_VERSION
 ARG GOLANG_VERSION
 COPY ./scripts/rh_linux_deps.sh /
 RUN CMAKE_VERSION=${CMAKE_VERSION} GOLANG_VERSION=${GOLANG_VERSION} sh /rh_linux_deps.sh
-ENV PATH /opt/rh/devtoolset-10/root/usr/bin:$PATH
-COPY --from=llm-code / /go/src/github.com/jmorganca/ollama/
-WORKDIR /go/src/github.com/jmorganca/ollama/llm/generate
-# Note, we only build the "base" CPU variant on arm since avx/avx2 are x86 features
+ENV PATH /opt/rh/gcc-toolset-10/root/usr/bin:$PATH
+COPY --from=llm-code / /go/src/github.com/ollama/ollama/
 ARG OLLAMA_CUSTOM_CPU_DEFS
 ARG CGO_CFLAGS
-RUN OLLAMA_CPU_TARGET="cpu" sh gen_linux.sh
+WORKDIR /go/src/github.com/ollama/ollama/llm/generate
+
+FROM --platform=linux/arm64 cpu-builder-arm64 AS static-build-arm64
+RUN OLLAMA_CPU_TARGET="static" sh gen_linux.sh
+FROM --platform=linux/arm64 cpu-builder-arm64 AS cpu-build-arm64
+RUN OLLAMA_SKIP_STATIC_GENERATE=1 OLLAMA_CPU_TARGET="cpu" sh gen_linux.sh
+
 
 # Intermediate stage used for ./scripts/build_linux.sh
 FROM --platform=linux/amd64 cpu-build-amd64 AS build-amd64
 ENV CGO_ENABLED 1
-WORKDIR /go/src/github.com/jmorganca/ollama
+WORKDIR /go/src/github.com/ollama/ollama
 COPY . .
-COPY --from=cpu_avx-build-amd64 /go/src/github.com/jmorganca/ollama/llm/llama.cpp/build/linux/ llm/llama.cpp/build/linux/
-COPY --from=cpu_avx2-build-amd64 /go/src/github.com/jmorganca/ollama/llm/llama.cpp/build/linux/ llm/llama.cpp/build/linux/
-COPY --from=cuda-build-amd64 /go/src/github.com/jmorganca/ollama/llm/llama.cpp/build/linux/ llm/llama.cpp/build/linux/
-COPY --from=rocm-5-build-amd64 /go/src/github.com/jmorganca/ollama/llm/llama.cpp/build/linux/ llm/llama.cpp/build/linux/
-COPY --from=rocm-6-build-amd64 /go/src/github.com/jmorganca/ollama/llm/llama.cpp/build/linux/ llm/llama.cpp/build/linux/
+COPY --from=static-build-amd64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
+COPY --from=cpu_avx-build-amd64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
+COPY --from=cpu_avx2-build-amd64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
+COPY --from=cuda-build-amd64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
+COPY --from=rocm-build-amd64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
+COPY --from=rocm-build-amd64 /go/src/github.com/ollama/ollama/dist/deps/ ./dist/deps/
 ARG GOFLAGS
 ARG CGO_CFLAGS
-RUN go build .
+RUN go build -trimpath .
 
 # Intermediate stage used for ./scripts/build_linux.sh
 FROM --platform=linux/arm64 cpu-build-arm64 AS build-arm64
 ENV CGO_ENABLED 1
 ARG GOLANG_VERSION
-WORKDIR /go/src/github.com/jmorganca/ollama
+WORKDIR /go/src/github.com/ollama/ollama
 COPY . .
-COPY --from=cuda-build-arm64 /go/src/github.com/jmorganca/ollama/llm/llama.cpp/build/linux/ llm/llama.cpp/build/linux/
+COPY --from=static-build-arm64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
+COPY --from=cuda-build-arm64 /go/src/github.com/ollama/ollama/llm/build/linux/ llm/build/linux/
 ARG GOFLAGS
 ARG CGO_CFLAGS
-RUN go build .
+RUN go build -trimpath .
 
 # Runtime stages
 FROM --platform=linux/amd64 ubuntu:22.04 as runtime-amd64
 RUN apt-get update && apt-get install -y ca-certificates
-COPY --from=build-amd64 /go/src/github.com/jmorganca/ollama/ollama /bin/ollama
+COPY --from=build-amd64 /go/src/github.com/ollama/ollama/ollama /bin/ollama
 FROM --platform=linux/arm64 ubuntu:22.04 as runtime-arm64
 RUN apt-get update && apt-get install -y ca-certificates
-COPY --from=build-arm64 /go/src/github.com/jmorganca/ollama/ollama /bin/ollama
+COPY --from=build-arm64 /go/src/github.com/ollama/ollama/ollama /bin/ollama
 
 # Radeon images are much larger so we keep it distinct from the CPU/CUDA image
-FROM --platform=linux/amd64 rocm/dev-centos-7:5.7.1-complete as runtime-rocm
+FROM --platform=linux/amd64 rocm/dev-centos-7:${ROCM_VERSION}-complete as runtime-rocm
 RUN update-pciids
-COPY --from=build-amd64 /go/src/github.com/jmorganca/ollama/ollama /bin/ollama
+COPY --from=build-amd64 /go/src/github.com/ollama/ollama/ollama /bin/ollama
 EXPOSE 11434
 ENV OLLAMA_HOST 0.0.0.0
 
@@ -132,6 +138,7 @@ ENV OLLAMA_HOST 0.0.0.0
 ENV PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
 ENV LD_LIBRARY_PATH=/usr/local/nvidia/lib:/usr/local/nvidia/lib64
 ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility
+ENV NVIDIA_VISIBLE_DEVICES=all
 
 ENTRYPOINT ["/bin/ollama"]
 CMD ["serve"]

README.md

@@ -1,12 +1,12 @@
 <div align="center">
-  <img alt="ollama" height="200px" src="https://github.com/jmorganca/ollama/assets/3325447/0d0b44e2-8f4a-4e99-9b52-a5c1c741c8f7">
+ <img alt="ollama" height="200px" src="https://github.com/ollama/ollama/assets/3325447/0d0b44e2-8f4a-4e99-9b52-a5c1c741c8f7">
 </div>
 
 # Ollama
 
 [![Discord](https://dcbadge.vercel.app/api/server/ollama?style=flat&compact=true)](https://discord.gg/ollama)
 
-Get up and running with large language models locally.
+Get up and running with large language models.
 
 ### macOS
 
@@ -22,7 +22,7 @@ Get up and running with large language models locally.
 curl -fsSL https://ollama.com/install.sh | sh
 ```
 
-[Manual install instructions](https://github.com/jmorganca/ollama/blob/main/docs/linux.md)
+[Manual install instructions](https://github.com/ollama/ollama/blob/main/docs/linux.md)
 
 ### Docker
 
@@ -35,10 +35,10 @@ The official [Ollama Docker image](https://hub.docker.com/r/ollama/ollama) `olla
 
 ## Quickstart
 
-To run and chat with [Llama 2](https://ollama.com/library/llama2):
+To run and chat with [Llama 3](https://ollama.com/library/llama3):
 
 ```
-ollama run llama2
+ollama run llama3
 ```
 
 ## Model library
@@ -49,19 +49,20 @@ Here are some example models that can be downloaded:
 
 | Model              | Parameters | Size  | Download                       |
 | ------------------ | ---------- | ----- | ------------------------------ |
-| Llama 2            | 7B         | 3.8GB | `ollama run llama2`            |
+| Llama 3            | 8B         | 4.7GB | `ollama run llama3`            |
+| Llama 3            | 70B        | 40GB  | `ollama run llama3:70b`        |
+| Phi 3 Mini         | 3.8B       | 2.3GB | `ollama run phi3`              |
+| Phi 3 Medium       | 14B        | 7.9GB | `ollama run phi3:medium`       |
+| Gemma 2            | 9B         | 5.5GB | `ollama run gemma2`            |
+| Gemma 2            | 27B        | 16GB  | `ollama run gemma2:27b`        |
 | Mistral            | 7B         | 4.1GB | `ollama run mistral`           |
-| Dolphin Phi        | 2.7B       | 1.6GB | `ollama run dolphin-phi`       |
-| Phi-2              | 2.7B       | 1.7GB | `ollama run phi`               |
+| Moondream 2        | 1.4B       | 829MB | `ollama run moondream`         |
 | Neural Chat        | 7B         | 4.1GB | `ollama run neural-chat`       |
 | Starling           | 7B         | 4.1GB | `ollama run starling-lm`       |
 | Code Llama         | 7B         | 3.8GB | `ollama run codellama`         |
 | Llama 2 Uncensored | 7B         | 3.8GB | `ollama run llama2-uncensored` |
-| Llama 2 13B        | 13B        | 7.3GB | `ollama run llama2:13b`        |
-| Llama 2 70B        | 70B        | 39GB  | `ollama run llama2:70b`        |
-| Orca Mini          | 3B         | 1.9GB | `ollama run orca-mini`         |
-| Vicuna             | 7B         | 3.8GB | `ollama run vicuna`            |
 | LLaVA              | 7B         | 4.5GB | `ollama run llava`             |
+| Solar              | 10.7B      | 6.1GB | `ollama run solar`             |
 
 > Note: You should have at least 8 GB of RAM available to run the 7B models, 16 GB to run the 13B models, and 32 GB to run the 33B models.
 
@@ -95,16 +96,16 @@ See the [guide](docs/import.md) on importing models for more information.
 
 ### Customize a prompt
 
-Models from the Ollama library can be customized with a prompt. For example, to customize the `llama2` model:
+Models from the Ollama library can be customized with a prompt. For example, to customize the `llama3` model:
 
 ```
-ollama pull llama2
+ollama pull llama3
 ```
 
 Create a `Modelfile`:
 
 ```
-FROM llama2
+FROM llama3
 
 # set the temperature to 1 [higher is more creative, lower is more coherent]
 PARAMETER temperature 1
@@ -139,7 +140,7 @@ ollama create mymodel -f ./Modelfile
 ### Pull a model
 
 ```
-ollama pull llama2
+ollama pull llama3
 ```
 
 > This command can also be used to update a local model. Only the diff will be pulled.
@@ -147,13 +148,13 @@ ollama pull llama2
 ### Remove a model
 
 ```
-ollama rm llama2
+ollama rm llama3
 ```
 
 ### Copy a model
 
 ```
-ollama cp llama2 my-llama2
+ollama cp llama3 my-model
 ```
 
 ### Multiline input
@@ -174,13 +175,19 @@ I'm a basic program that prints the famous "Hello, world!" message to the consol
 The image features a yellow smiley face, which is likely the central focus of the picture.
 ```
 
-### Pass in prompt as arguments
+### Pass the prompt as an argument
 
 ```
-$ ollama run llama2 "Summarize this file: $(cat README.md)"
+$ ollama run llama3 "Summarize this file: $(cat README.md)"
  Ollama is a lightweight, extensible framework for building and running language models on the local machine. It provides a simple API for creating, running, and managing models, as well as a library of pre-built models that can be easily used in a variety of applications.
 ```
 
+### Show model information
+
+```
+ollama show llama3
+```
+
 ### List models on your computer
 
 ```
@@ -193,25 +200,7 @@ ollama list
 
 ## Building
 
-Install `cmake` and `go`:
-
-```
-brew install cmake go
-```
-
-Then generate dependencies:
-
-```
-go generate ./...
-```
-
-Then build the binary:
-
-```
-go build .
-```
-
-More detailed instructions can be found in the [developer guide](https://github.com/jmorganca/ollama/blob/main/docs/development.md)
+See the [developer guide](https://github.com/ollama/ollama/blob/main/docs/development.md)
 
 ### Running local builds
 
@@ -224,7 +213,7 @@ Next, start the server:
 Finally, in a separate shell, run a model:
 
 ```
-./ollama run llama2
+./ollama run llama3
 ```
 
 ## REST API
@@ -235,7 +224,7 @@ Ollama has a REST API for running and managing models.
 
 ```
 curl http://localhost:11434/api/generate -d '{
-  "model": "llama2",
+  "model": "llama3",
   "prompt":"Why is the sky blue?"
 }'
 ```
@@ -244,7 +233,7 @@ curl http://localhost:11434/api/generate -d '{
 
 ```
 curl http://localhost:11434/api/chat -d '{
-  "model": "mistral",
+  "model": "llama3",
   "messages": [
     { "role": "user", "content": "why is the sky blue?" }
   ]
@@ -257,20 +246,56 @@ See the [API documentation](./docs/api.md) for all endpoints.
 
 ### Web & Desktop
 
+- [Open WebUI](https://github.com/open-webui/open-webui)
+- [Enchanted (macOS native)](https://github.com/AugustDev/enchanted)
+- [Hollama](https://github.com/fmaclen/hollama)
+- [Lollms-Webui](https://github.com/ParisNeo/lollms-webui)
+- [LibreChat](https://github.com/danny-avila/LibreChat)
 - [Bionic GPT](https://github.com/bionic-gpt/bionic-gpt)
 - [HTML UI](https://github.com/rtcfirefly/ollama-ui)
+- [Saddle](https://github.com/jikkuatwork/saddle)
 - [Chatbot UI](https://github.com/ivanfioravanti/chatbot-ollama)
+- [Chatbot UI v2](https://github.com/mckaywrigley/chatbot-ui)
 - [Typescript UI](https://github.com/ollama-interface/Ollama-Gui?tab=readme-ov-file)
 - [Minimalistic React UI for Ollama Models](https://github.com/richawo/minimal-llm-ui)
-- [Open WebUI](https://github.com/open-webui/open-webui)
 - [Ollamac](https://github.com/kevinhermawan/Ollamac)
-- [big-AGI](https://github.com/enricoros/big-agi/blob/main/docs/config-ollama.md)
+- [big-AGI](https://github.com/enricoros/big-AGI/blob/main/docs/config-local-ollama.md)
 - [Cheshire Cat assistant framework](https://github.com/cheshire-cat-ai/core)
 - [Amica](https://github.com/semperai/amica)
 - [chatd](https://github.com/BruceMacD/chatd)
 - [Ollama-SwiftUI](https://github.com/kghandour/Ollama-SwiftUI)
+- [Dify.AI](https://github.com/langgenius/dify)
 - [MindMac](https://mindmac.app)
 - [NextJS Web Interface for Ollama](https://github.com/jakobhoeg/nextjs-ollama-llm-ui)
+- [Msty](https://msty.app)
+- [Chatbox](https://github.com/Bin-Huang/Chatbox)
+- [WinForm Ollama Copilot](https://github.com/tgraupmann/WinForm_Ollama_Copilot)
+- [NextChat](https://github.com/ChatGPTNextWeb/ChatGPT-Next-Web) with [Get Started Doc](https://docs.nextchat.dev/models/ollama)
+- [Alpaca WebUI](https://github.com/mmo80/alpaca-webui)
+- [OllamaGUI](https://github.com/enoch1118/ollamaGUI)
+- [OpenAOE](https://github.com/InternLM/OpenAOE)
+- [Odin Runes](https://github.com/leonid20000/OdinRunes)
+- [LLM-X](https://github.com/mrdjohnson/llm-x) (Progressive Web App)
+- [AnythingLLM (Docker + MacOs/Windows/Linux native app)](https://github.com/Mintplex-Labs/anything-llm)
+- [Ollama Basic Chat: Uses HyperDiv Reactive UI](https://github.com/rapidarchitect/ollama_basic_chat)
+- [Ollama-chats RPG](https://github.com/drazdra/ollama-chats)
+- [QA-Pilot](https://github.com/reid41/QA-Pilot) (Chat with Code Repository)
+- [ChatOllama](https://github.com/sugarforever/chat-ollama) (Open Source Chatbot based on Ollama with Knowledge Bases)
+- [CRAG Ollama Chat](https://github.com/Nagi-ovo/CRAG-Ollama-Chat) (Simple Web Search with Corrective RAG)
+- [RAGFlow](https://github.com/infiniflow/ragflow) (Open-source Retrieval-Augmented Generation engine based on deep document understanding)
+- [StreamDeploy](https://github.com/StreamDeploy-DevRel/streamdeploy-llm-app-scaffold) (LLM Application Scaffold)
+- [chat](https://github.com/swuecho/chat) (chat web app for teams)
+- [Lobe Chat](https://github.com/lobehub/lobe-chat) with [Integrating Doc](https://lobehub.com/docs/self-hosting/examples/ollama)
+- [Ollama RAG Chatbot](https://github.com/datvodinh/rag-chatbot.git) (Local Chat with multiple PDFs using Ollama and RAG)
+- [BrainSoup](https://www.nurgo-software.com/products/brainsoup) (Flexible native client with RAG & multi-agent automation)
+- [macai](https://github.com/Renset/macai) (macOS client for Ollama, ChatGPT, and other compatible API back-ends)
+- [Olpaka](https://github.com/Otacon/olpaka) (User-friendly Flutter Web App for Ollama)
+- [OllamaSpring](https://github.com/CrazyNeil/OllamaSpring) (Ollama Client for macOS)
+- [LLocal.in](https://github.com/kartikm7/llocal) (Easy to use Electron Desktop Client for Ollama)
+- [Ollama with Google Mesop](https://github.com/rapidarchitect/ollama_mesop/) (Mesop Chat Client implementation with Ollama)
+- [Kerlig AI](https://www.kerlig.com/) (AI writing assistant for macOS)
+- [AI Studio](https://github.com/MindWorkAI/AI-Studio)
+- [Sidellama](https://github.com/gyopak/sidellama) (browser-based LLM client)
 
 ### Terminal
 
@@ -279,34 +304,45 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [Emacs client](https://github.com/zweifisch/ollama)
 - [gen.nvim](https://github.com/David-Kunz/gen.nvim)
 - [ollama.nvim](https://github.com/nomnivore/ollama.nvim)
+- [ollero.nvim](https://github.com/marco-souza/ollero.nvim)
 - [ollama-chat.nvim](https://github.com/gerazov/ollama-chat.nvim)
 - [ogpt.nvim](https://github.com/huynle/ogpt.nvim)
 - [gptel Emacs client](https://github.com/karthink/gptel)
 - [Oatmeal](https://github.com/dustinblackman/oatmeal)
 - [cmdh](https://github.com/pgibler/cmdh)
+- [ooo](https://github.com/npahlfer/ooo)
+- [shell-pilot](https://github.com/reid41/shell-pilot)
 - [tenere](https://github.com/pythops/tenere)
 - [llm-ollama](https://github.com/taketwo/llm-ollama) for [Datasette's LLM CLI](https://llm.datasette.io/en/stable/).
+- [typechat-cli](https://github.com/anaisbetts/typechat-cli)
 - [ShellOracle](https://github.com/djcopley/ShellOracle)
+- [tlm](https://github.com/yusufcanb/tlm)
+- [podman-ollama](https://github.com/ericcurtin/podman-ollama)
+- [gollama](https://github.com/sammcj/gollama)
 
 ### Database
 
-- [MindsDB](https://github.com/mindsdb/mindsdb/blob/staging/mindsdb/integrations/handlers/ollama_handler/README.md)
+- [MindsDB](https://github.com/mindsdb/mindsdb/blob/staging/mindsdb/integrations/handlers/ollama_handler/README.md) (Connects Ollama models with nearly 200 data platforms and apps)
+- [chromem-go](https://github.com/philippgille/chromem-go/blob/v0.5.0/embed_ollama.go) with [example](https://github.com/philippgille/chromem-go/tree/v0.5.0/examples/rag-wikipedia-ollama)
 
 ### Package managers
 
 - [Pacman](https://archlinux.org/packages/extra/x86_64/ollama/)
 - [Helm Chart](https://artifacthub.io/packages/helm/ollama-helm/ollama)
+- [Guix channel](https://codeberg.org/tusharhero/ollama-guix)
 
 ### Libraries
 
 - [LangChain](https://python.langchain.com/docs/integrations/llms/ollama) and [LangChain.js](https://js.langchain.com/docs/modules/model_io/models/llms/integrations/ollama) with [example](https://js.langchain.com/docs/use_cases/question_answering/local_retrieval_qa)
 - [LangChainGo](https://github.com/tmc/langchaingo/) with [example](https://github.com/tmc/langchaingo/tree/main/examples/ollama-completion-example)
+- [LangChain4j](https://github.com/langchain4j/langchain4j) with [example](https://github.com/langchain4j/langchain4j-examples/tree/main/ollama-examples/src/main/java)
+- [LangChainRust](https://github.com/Abraxas-365/langchain-rust) with [example](https://github.com/Abraxas-365/langchain-rust/blob/main/examples/llm_ollama.rs)
 - [LlamaIndex](https://gpt-index.readthedocs.io/en/stable/examples/llm/ollama.html)
-- [LangChain4j](https://github.com/langchain4j/langchain4j/tree/main/langchain4j-ollama)
 - [LiteLLM](https://github.com/BerriAI/litellm)
 - [OllamaSharp for .NET](https://github.com/awaescher/OllamaSharp)
 - [Ollama for Ruby](https://github.com/gbaptista/ollama-ai)
 - [Ollama-rs for Rust](https://github.com/pepperoni21/ollama-rs)
+- [Ollama-hpp for C++](https://github.com/jmont-dev/ollama-hpp)
 - [Ollama4j for Java](https://github.com/amithkoujalgi/ollama4j)
 - [ModelFusion Typescript Library](https://modelfusion.dev/integration/model-provider/ollama)
 - [OllamaKit for Swift](https://github.com/kevinhermawan/OllamaKit)
@@ -315,8 +351,15 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [LangChainDart](https://github.com/davidmigloz/langchain_dart)
 - [Semantic Kernel - Python](https://github.com/microsoft/semantic-kernel/tree/main/python/semantic_kernel/connectors/ai/ollama)
 - [Haystack](https://github.com/deepset-ai/haystack-integrations/blob/main/integrations/ollama.md)
+- [Elixir LangChain](https://github.com/brainlid/langchain)
 - [Ollama for R - rollama](https://github.com/JBGruber/rollama)
+- [Ollama for R - ollama-r](https://github.com/hauselin/ollama-r)
 - [Ollama-ex for Elixir](https://github.com/lebrunel/ollama-ex)
+- [Ollama Connector for SAP ABAP](https://github.com/b-tocs/abap_btocs_ollama)
+- [Testcontainers](https://testcontainers.com/modules/ollama/)
+- [Portkey](https://portkey.ai/docs/welcome/integration-guides/ollama)
+- [PromptingTools.jl](https://github.com/svilupp/PromptingTools.jl) with an [example](https://svilupp.github.io/PromptingTools.jl/dev/examples/working_with_ollama)
+- [LlamaScript](https://github.com/Project-Llama/llamascript)
 
 ### Mobile
 
@@ -330,13 +373,29 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [Continue](https://github.com/continuedev/continue)
 - [Obsidian Ollama plugin](https://github.com/hinterdupfinger/obsidian-ollama)
 - [Logseq Ollama plugin](https://github.com/omagdy7/ollama-logseq)
+- [NotesOllama](https://github.com/andersrex/notesollama) (Apple Notes Ollama plugin)
 - [Dagger Chatbot](https://github.com/samalba/dagger-chatbot)
 - [Discord AI Bot](https://github.com/mekb-turtle/discord-ai-bot)
 - [Ollama Telegram Bot](https://github.com/ruecat/ollama-telegram)
 - [Hass Ollama Conversation](https://github.com/ej52/hass-ollama-conversation)
 - [Rivet plugin](https://github.com/abrenneke/rivet-plugin-ollama)
-- [Llama Coder](https://github.com/ex3ndr/llama-coder) (Copilot alternative using Ollama)
 - [Obsidian BMO Chatbot plugin](https://github.com/longy2k/obsidian-bmo-chatbot)
+- [Cliobot](https://github.com/herval/cliobot) (Telegram bot with Ollama support)
+- [Copilot for Obsidian plugin](https://github.com/logancyang/obsidian-copilot)
+- [Obsidian Local GPT plugin](https://github.com/pfrankov/obsidian-local-gpt)
 - [Open Interpreter](https://docs.openinterpreter.com/language-model-setup/local-models/ollama)
+- [Llama Coder](https://github.com/ex3ndr/llama-coder) (Copilot alternative using Ollama)
+- [Ollama Copilot](https://github.com/bernardo-bruning/ollama-copilot) (Proxy that allows you to use ollama as a copilot like Github copilot)
 - [twinny](https://github.com/rjmacarthy/twinny) (Copilot and Copilot chat alternative using Ollama)
 - [Wingman-AI](https://github.com/RussellCanfield/wingman-ai) (Copilot code and chat alternative using Ollama and HuggingFace)
+- [Page Assist](https://github.com/n4ze3m/page-assist) (Chrome Extension)
+- [AI Telegram Bot](https://github.com/tusharhero/aitelegrambot) (Telegram bot using Ollama in backend)
+- [AI ST Completion](https://github.com/yaroslavyaroslav/OpenAI-sublime-text) (Sublime Text 4 AI assistant plugin with Ollama support)
+- [Discord-Ollama Chat Bot](https://github.com/kevinthedang/discord-ollama) (Generalized TypeScript Discord Bot w/ Tuning Documentation)
+- [Discord AI chat/moderation bot](https://github.com/rapmd73/Companion) Chat/moderation bot written in python. Uses Ollama to create personalities.
+- [Headless Ollama](https://github.com/nischalj10/headless-ollama) (Scripts to automatically install ollama client & models on any OS for apps that depends on ollama server)
+
+### Supported backends
+
+- [llama.cpp](https://github.com/ggerganov/llama.cpp) project founded by Georgi Gerganov.
+

api/client.go

@@ -1,3 +1,16 @@
+// Package api implements the client-side API for code wishing to interact
+// with the ollama service. The methods of the [Client] type correspond to
+// the ollama REST API as described in [the API documentation].
+// The ollama command-line client itself uses this package to interact with
+// the backend service.
+//
+// # Examples
+//
+// Several examples of using this package are available [in the GitHub
+// repository].
+//
+// [the API documentation]: https://github.com/ollama/ollama/blob/main/docs/api.md
+// [in the GitHub repository]: https://github.com/ollama/ollama/tree/main/examples
 package api
 
 import (
@@ -5,23 +18,23 @@ import (
 	"bytes"
 	"context"
 	"encoding/json"
-	"errors"
 	"fmt"
 	"io"
 	"net"
 	"net/http"
 	"net/url"
-	"os"
 	"runtime"
-	"strings"
 
-	"github.com/jmorganca/ollama/format"
-	"github.com/jmorganca/ollama/version"
+	"github.com/ollama/ollama/envconfig"
+	"github.com/ollama/ollama/format"
+	"github.com/ollama/ollama/version"
 )
 
+// Client encapsulates client state for interacting with the ollama
+// service. Use [ClientFromEnvironment] to create new Clients.
 type Client struct {
 	base *url.URL
-	http http.Client
+	http *http.Client
 }
 
 func checkError(resp *http.Response, body []byte) error {
@@ -40,56 +53,32 @@ func checkError(resp *http.Response, body []byte) error {
 	return apiError
 }
 
+// ClientFromEnvironment creates a new [Client] using configuration from the
+// environment variable OLLAMA_HOST, which points to the network host and
+// port on which the ollama service is listenting. The format of this variable
+// is:
+//
+//	<scheme>://<host>:<port>
+//
+// If the variable is not specified, a default ollama host and port will be
+// used.
 func ClientFromEnvironment() (*Client, error) {
-	defaultPort := "11434"
+	ollamaHost := envconfig.Host
 
-	scheme, hostport, ok := strings.Cut(os.Getenv("OLLAMA_HOST"), "://")
-	switch {
-	case !ok:
-		scheme, hostport = "http", os.Getenv("OLLAMA_HOST")
-	case scheme == "http":
-		defaultPort = "80"
-	case scheme == "https":
-		defaultPort = "443"
-	}
-
-	// trim trailing slashes
-	hostport = strings.TrimRight(hostport, "/")
-
-	host, port, err := net.SplitHostPort(hostport)
-	if err != nil {
-		host, port = "127.0.0.1", defaultPort
-		if ip := net.ParseIP(strings.Trim(hostport, "[]")); ip != nil {
-			host = ip.String()
-		} else if hostport != "" {
-			host = hostport
-		}
-	}
-
-	client := Client{
+	return &Client{
 		base: &url.URL{
-			Scheme: scheme,
-			Host:   net.JoinHostPort(host, port),
+			Scheme: ollamaHost.Scheme,
+			Host:   net.JoinHostPort(ollamaHost.Host, ollamaHost.Port),
 		},
-	}
+		http: http.DefaultClient,
+	}, nil
+}
 
-	mockRequest, err := http.NewRequest(http.MethodHead, client.base.String(), nil)
-	if err != nil {
-		return nil, err
+func NewClient(base *url.URL, http *http.Client) *Client {
+	return &Client{
+		base: base,
+		http: http,
 	}
-
-	proxyURL, err := http.ProxyFromEnvironment(mockRequest)
-	if err != nil {
-		return nil, err
-	}
-
-	client.http = http.Client{
-		Transport: &http.Transport{
-			Proxy: http.ProxyURL(proxyURL),
-		},
-	}
-
-	return &client, nil
 }
 
 func (c *Client) do(ctx context.Context, method, path string, reqData, respData any) error {
@@ -208,8 +197,14 @@ func (c *Client) stream(ctx context.Context, method, path string, data any, fn f
 	return nil
 }
 
+// GenerateResponseFunc is a function that [Client.Generate] invokes every time
+// a response is received from the service. If this function returns an error,
+// [Client.Generate] will stop generating and return this error.
 type GenerateResponseFunc func(GenerateResponse) error
 
+// Generate generates a response for a given prompt. The req parameter should
+// be populated with prompt details. fn is called for each response (there may
+// be multiple responses, e.g. in case streaming is enabled).
 func (c *Client) Generate(ctx context.Context, req *GenerateRequest, fn GenerateResponseFunc) error {
 	return c.stream(ctx, http.MethodPost, "/api/generate", req, func(bts []byte) error {
 		var resp GenerateResponse
@@ -221,8 +216,15 @@ func (c *Client) Generate(ctx context.Context, req *GenerateRequest, fn Generate
 	})
 }
 
+// ChatResponseFunc is a function that [Client.Chat] invokes every time
+// a response is received from the service. If this function returns an error,
+// [Client.Chat] will stop generating and return this error.
 type ChatResponseFunc func(ChatResponse) error
 
+// Chat generates the next message in a chat. [ChatRequest] may contain a
+// sequence of messages which can be used to maintain chat history with a model.
+// fn is called for each response (there may be multiple responses, e.g. if case
+// streaming is enabled).
 func (c *Client) Chat(ctx context.Context, req *ChatRequest, fn ChatResponseFunc) error {
 	return c.stream(ctx, http.MethodPost, "/api/chat", req, func(bts []byte) error {
 		var resp ChatResponse
@@ -234,8 +236,14 @@ func (c *Client) Chat(ctx context.Context, req *ChatRequest, fn ChatResponseFunc
 	})
 }
 
+// PullProgressFunc is a function that [Client.Pull] invokes every time there
+// is progress with a "pull" request sent to the service. If this function
+// returns an error, [Client.Pull] will stop the process and return this error.
 type PullProgressFunc func(ProgressResponse) error
 
+// Pull downloads a model from the ollama library. fn is called each time
+// progress is made on the request and can be used to display a progress bar,
+// etc.
 func (c *Client) Pull(ctx context.Context, req *PullRequest, fn PullProgressFunc) error {
 	return c.stream(ctx, http.MethodPost, "/api/pull", req, func(bts []byte) error {
 		var resp ProgressResponse
@@ -247,8 +255,14 @@ func (c *Client) Pull(ctx context.Context, req *PullRequest, fn PullProgressFunc
 	})
 }
 
+// PushProgressFunc is a function that [Client.Push] invokes when progress is
+// made.
+// It's similar to other progress function types like [PullProgressFunc].
 type PushProgressFunc func(ProgressResponse) error
 
+// Push uploads a model to the model library; requires registering for ollama.ai
+// and adding a public key first. fn is called each time progress is made on
+// the request and can be used to display a progress bar, etc.
 func (c *Client) Push(ctx context.Context, req *PushRequest, fn PushProgressFunc) error {
 	return c.stream(ctx, http.MethodPost, "/api/push", req, func(bts []byte) error {
 		var resp ProgressResponse
@@ -260,8 +274,15 @@ func (c *Client) Push(ctx context.Context, req *PushRequest, fn PushProgressFunc
 	})
 }
 
+// CreateProgressFunc is a function that [Client.Create] invokes when progress
+// is made.
+// It's similar to other progress function types like [PullProgressFunc].
 type CreateProgressFunc func(ProgressResponse) error
 
+// Create creates a model from a [Modelfile]. fn is a progress function that
+// behaves similarly to other methods (see [Client.Pull]).
+//
+// [Modelfile]: https://github.com/ollama/ollama/blob/main/docs/modelfile.md
 func (c *Client) Create(ctx context.Context, req *CreateRequest, fn CreateProgressFunc) error {
 	return c.stream(ctx, http.MethodPost, "/api/create", req, func(bts []byte) error {
 		var resp ProgressResponse
@@ -273,6 +294,7 @@ func (c *Client) Create(ctx context.Context, req *CreateRequest, fn CreateProgre
 	})
 }
 
+// List lists models that are available locally.
 func (c *Client) List(ctx context.Context) (*ListResponse, error) {
 	var lr ListResponse
 	if err := c.do(ctx, http.MethodGet, "/api/tags", nil, &lr); err != nil {
@@ -281,6 +303,17 @@ func (c *Client) List(ctx context.Context) (*ListResponse, error) {
 	return &lr, nil
 }
 
+// List running models.
+func (c *Client) ListRunning(ctx context.Context) (*ProcessResponse, error) {
+	var lr ProcessResponse
+	if err := c.do(ctx, http.MethodGet, "/api/ps", nil, &lr); err != nil {
+		return nil, err
+	}
+	return &lr, nil
+}
+
+// Copy copies a model - creating a model with another name from an existing
+// model.
 func (c *Client) Copy(ctx context.Context, req *CopyRequest) error {
 	if err := c.do(ctx, http.MethodPost, "/api/copy", req, nil); err != nil {
 		return err
@@ -288,6 +321,7 @@ func (c *Client) Copy(ctx context.Context, req *CopyRequest) error {
 	return nil
 }
 
+// Delete deletes a model and its data.
 func (c *Client) Delete(ctx context.Context, req *DeleteRequest) error {
 	if err := c.do(ctx, http.MethodDelete, "/api/delete", req, nil); err != nil {
 		return err
@@ -295,6 +329,7 @@ func (c *Client) Delete(ctx context.Context, req *DeleteRequest) error {
 	return nil
 }
 
+// Show obtains model information, including details, modelfile, license etc.
 func (c *Client) Show(ctx context.Context, req *ShowRequest) (*ShowResponse, error) {
 	var resp ShowResponse
 	if err := c.do(ctx, http.MethodPost, "/api/show", req, &resp); err != nil {
@@ -303,12 +338,25 @@ func (c *Client) Show(ctx context.Context, req *ShowRequest) (*ShowResponse, err
 	return &resp, nil
 }
 
+// Hearbeat checks if the server has started and is responsive; if yes, it
+// returns nil, otherwise an error.
 func (c *Client) Heartbeat(ctx context.Context) error {
 	if err := c.do(ctx, http.MethodHead, "/", nil, nil); err != nil {
 		return err
 	}
 	return nil
 }
+
+// Embed generates embeddings from a model.
+func (c *Client) Embed(ctx context.Context, req *EmbedRequest) (*EmbedResponse, error) {
+	var resp EmbedResponse
+	if err := c.do(ctx, http.MethodPost, "/api/embed", req, &resp); err != nil {
+		return nil, err
+	}
+	return &resp, nil
+}
+
+// Embeddings generates an embedding from a model.
 func (c *Client) Embeddings(ctx context.Context, req *EmbeddingRequest) (*EmbeddingResponse, error) {
 	var resp EmbeddingResponse
 	if err := c.do(ctx, http.MethodPost, "/api/embeddings", req, &resp); err != nil {
@@ -317,21 +365,13 @@ func (c *Client) Embeddings(ctx context.Context, req *EmbeddingRequest) (*Embedd
 	return &resp, nil
 }
 
+// CreateBlob creates a blob from a file on the server. digest is the
+// expected SHA256 digest of the file, and r represents the file.
 func (c *Client) CreateBlob(ctx context.Context, digest string, r io.Reader) error {
-	if err := c.do(ctx, http.MethodHead, fmt.Sprintf("/api/blobs/%s", digest), nil, nil); err != nil {
-		var statusError StatusError
-		if !errors.As(err, &statusError) || statusError.StatusCode != http.StatusNotFound {
-			return err
-		}
-
-		if err := c.do(ctx, http.MethodPost, fmt.Sprintf("/api/blobs/%s", digest), r, nil); err != nil {
-			return err
-		}
-	}
-
-	return nil
+	return c.do(ctx, http.MethodPost, fmt.Sprintf("/api/blobs/%s", digest), r, nil)
 }
 
+// Version returns the Ollama server version as a string.
 func (c *Client) Version(ctx context.Context) (string, error) {
 	var version struct {
 		Version string `json:"version"`

api/client_test.go

@@ -1,6 +1,10 @@
 package api
 
-import "testing"
+import (
+	"testing"
+
+	"github.com/ollama/ollama/envconfig"
+)
 
 func TestClientFromEnvironment(t *testing.T) {
 	type testCase struct {
@@ -29,6 +33,7 @@ func TestClientFromEnvironment(t *testing.T) {
 	for k, v := range testCases {
 		t.Run(k, func(t *testing.T) {
 			t.Setenv("OLLAMA_HOST", v.value)
+			envconfig.LoadConfig()
 
 			client, err := ClientFromEnvironment()
 			if err != v.err {

api/types.go

@@ -3,6 +3,7 @@ package api
 import (
 	"encoding/json"
 	"fmt"
+	"log/slog"
 	"math"
 	"os"
 	"reflect"
@@ -11,6 +12,7 @@ import (
 	"time"
 )
 
+// StatusError is an error with and HTTP status code.
 type StatusError struct {
 	StatusCode   int
 	Status       string
@@ -31,43 +33,156 @@ func (e StatusError) Error() string {
 	}
 }
 
+// ImageData represents the raw binary data of an image file.
 type ImageData []byte
 
+// GenerateRequest describes a request sent by [Client.Generate]. While you
+// have to specify the Model and Prompt fields, all the other fields have
+// reasonable defaults for basic uses.
 type GenerateRequest struct {
-	Model     string      `json:"model"`
-	Prompt    string      `json:"prompt"`
-	System    string      `json:"system"`
-	Template  string      `json:"template"`
-	Context   []int       `json:"context,omitempty"`
-	Stream    *bool       `json:"stream,omitempty"`
-	Raw       bool        `json:"raw,omitempty"`
-	Format    string      `json:"format"`
-	KeepAlive *Duration   `json:"keep_alive,omitempty"`
-	Images    []ImageData `json:"images,omitempty"`
+	// Model is the model name; it should be a name familiar to Ollama from
+	// the library at https://ollama.com/library
+	Model string `json:"model"`
 
-	Options map[string]interface{} `json:"options"`
-}
+	// Prompt is the textual prompt to send to the model.
+	Prompt string `json:"prompt"`
 
-type ChatRequest struct {
-	Model     string    `json:"model"`
-	Messages  []Message `json:"messages"`
-	Stream    *bool     `json:"stream,omitempty"`
-	Format    string    `json:"format"`
+	// Suffix is the text that comes after the inserted text.
+	Suffix string `json:"suffix"`
+
+	// System overrides the model's default system message/prompt.
+	System string `json:"system"`
+
+	// Template overrides the model's default prompt template.
+	Template string `json:"template"`
+
+	// Context is the context parameter returned from a previous call to
+	// Generate call. It can be used to keep a short conversational memory.
+	Context []int `json:"context,omitempty"`
+
+	// Stream specifies whether the response is streaming; it is true by default.
+	Stream *bool `json:"stream,omitempty"`
+
+	// Raw set to true means that no formatting will be applied to the prompt.
+	Raw bool `json:"raw,omitempty"`
+
+	// Format specifies the format to return a response in.
+	Format string `json:"format"`
+
+	// KeepAlive controls how long the model will stay loaded in memory following
+	// this request.
 	KeepAlive *Duration `json:"keep_alive,omitempty"`
 
+	// Images is an optional list of base64-encoded images accompanying this
+	// request, for multimodal models.
+	Images []ImageData `json:"images,omitempty"`
+
+	// Options lists model-specific options. For example, temperature can be
+	// set through this field, if the model supports it.
 	Options map[string]interface{} `json:"options"`
 }
 
-type Message struct {
-	Role    string      `json:"role"` // one of ["system", "user", "assistant"]
-	Content string      `json:"content"`
-	Images  []ImageData `json:"images,omitempty"`
+// ChatRequest describes a request sent by [Client.Chat].
+type ChatRequest struct {
+	// Model is the model name, as in [GenerateRequest].
+	Model string `json:"model"`
+
+	// Messages is the messages of the chat - can be used to keep a chat memory.
+	Messages []Message `json:"messages"`
+
+	// Stream enable streaming of returned response; true by default.
+	Stream *bool `json:"stream,omitempty"`
+
+	// Format is the format to return the response in (e.g. "json").
+	Format string `json:"format"`
+
+	// KeepAlive controls how long the model will stay loaded into memory
+	// followin the request.
+	KeepAlive *Duration `json:"keep_alive,omitempty"`
+
+	// Tools is an optional list of tools the model has access to.
+	Tools `json:"tools,omitempty"`
+
+	// Options lists model-specific options.
+	Options map[string]interface{} `json:"options"`
 }
 
+type Tools []Tool
+
+func (t Tools) String() string {
+	bts, _ := json.Marshal(t)
+	return string(bts)
+}
+
+// Message is a single message in a chat sequence. The message contains the
+// role ("system", "user", or "assistant"), the content and an optional list
+// of images.
+type Message struct {
+	Role      string      `json:"role"`
+	Content   string      `json:"content"`
+	Images    []ImageData `json:"images,omitempty"`
+	ToolCalls []ToolCall  `json:"tool_calls,omitempty"`
+}
+
+func (m *Message) UnmarshalJSON(b []byte) error {
+	type Alias Message
+	var a Alias
+	if err := json.Unmarshal(b, &a); err != nil {
+		return err
+	}
+
+	*m = Message(a)
+	m.Role = strings.ToLower(m.Role)
+	return nil
+}
+
+type ToolCall struct {
+	Function ToolCallFunction `json:"function"`
+}
+
+type ToolCallFunction struct {
+	Name      string                    `json:"name"`
+	Arguments ToolCallFunctionArguments `json:"arguments"`
+}
+
+type ToolCallFunctionArguments map[string]any
+
+func (t *ToolCallFunctionArguments) String() string {
+	bts, _ := json.Marshal(t)
+	return string(bts)
+}
+
+type Tool struct {
+	Type     string       `json:"type"`
+	Function ToolFunction `json:"function"`
+}
+
+type ToolFunction struct {
+	Name        string `json:"name"`
+	Description string `json:"description"`
+	Parameters  struct {
+		Type       string   `json:"type"`
+		Required   []string `json:"required"`
+		Properties map[string]struct {
+			Type        string   `json:"type"`
+			Description string   `json:"description"`
+			Enum        []string `json:"enum,omitempty"`
+		} `json:"properties"`
+	} `json:"parameters"`
+}
+
+func (t *ToolFunction) String() string {
+	bts, _ := json.Marshal(t)
+	return string(bts)
+}
+
+// ChatResponse is the response returned by [Client.Chat]. Its fields are
+// similar to [GenerateResponse].
 type ChatResponse struct {
-	Model     string    `json:"model"`
-	CreatedAt time.Time `json:"created_at"`
-	Message   Message   `json:"message"`
+	Model      string    `json:"model"`
+	CreatedAt  time.Time `json:"created_at"`
+	Message    Message   `json:"message"`
+	DoneReason string    `json:"done_reason,omitempty"`
 
 	Done bool `json:"done"`
 
@@ -83,7 +198,8 @@ type Metrics struct {
 	EvalDuration       time.Duration `json:"eval_duration,omitempty"`
 }
 
-// Options specfied in GenerateRequest, if you add a new option here add it to the API docs also
+// Options specified in [GenerateRequest], if you add a new option here add it
+// to the API docs also.
 type Options struct {
 	Runner
 
@@ -109,46 +225,81 @@ type Options struct {
 
 // Runner options which must be set when the model is loaded into memory
 type Runner struct {
-	UseNUMA            bool    `json:"numa,omitempty"`
-	NumCtx             int     `json:"num_ctx,omitempty"`
-	NumBatch           int     `json:"num_batch,omitempty"`
-	NumGQA             int     `json:"num_gqa,omitempty"`
-	NumGPU             int     `json:"num_gpu,omitempty"`
-	MainGPU            int     `json:"main_gpu,omitempty"`
-	LowVRAM            bool    `json:"low_vram,omitempty"`
-	F16KV              bool    `json:"f16_kv,omitempty"`
-	LogitsAll          bool    `json:"logits_all,omitempty"`
-	VocabOnly          bool    `json:"vocab_only,omitempty"`
-	UseMMap            bool    `json:"use_mmap,omitempty"`
-	UseMLock           bool    `json:"use_mlock,omitempty"`
-	EmbeddingOnly      bool    `json:"embedding_only,omitempty"`
-	RopeFrequencyBase  float32 `json:"rope_frequency_base,omitempty"`
-	RopeFrequencyScale float32 `json:"rope_frequency_scale,omitempty"`
-	NumThread          int     `json:"num_thread,omitempty"`
+	UseNUMA   bool  `json:"numa,omitempty"`
+	NumCtx    int   `json:"num_ctx,omitempty"`
+	NumBatch  int   `json:"num_batch,omitempty"`
+	NumGPU    int   `json:"num_gpu,omitempty"`
+	MainGPU   int   `json:"main_gpu,omitempty"`
+	LowVRAM   bool  `json:"low_vram,omitempty"`
+	F16KV     bool  `json:"f16_kv,omitempty"`
+	LogitsAll bool  `json:"logits_all,omitempty"`
+	VocabOnly bool  `json:"vocab_only,omitempty"`
+	UseMMap   *bool `json:"use_mmap,omitempty"`
+	UseMLock  bool  `json:"use_mlock,omitempty"`
+	NumThread int   `json:"num_thread,omitempty"`
 }
 
-type EmbeddingRequest struct {
-	Model     string    `json:"model"`
-	Prompt    string    `json:"prompt"`
+// EmbedRequest is the request passed to [Client.Embed].
+type EmbedRequest struct {
+	// Model is the model name.
+	Model string `json:"model"`
+
+	// Input is the input to embed.
+	Input any `json:"input"`
+
+	// KeepAlive controls how long the model will stay loaded in memory following
+	// this request.
 	KeepAlive *Duration `json:"keep_alive,omitempty"`
 
+	Truncate *bool `json:"truncate,omitempty"`
+
+	// Options lists model-specific options.
 	Options map[string]interface{} `json:"options"`
 }
 
+// EmbedResponse is the response from [Client.Embed].
+type EmbedResponse struct {
+	Model      string      `json:"model"`
+	Embeddings [][]float32 `json:"embeddings"`
+}
+
+// EmbeddingRequest is the request passed to [Client.Embeddings].
+type EmbeddingRequest struct {
+	// Model is the model name.
+	Model string `json:"model"`
+
+	// Prompt is the textual prompt to embed.
+	Prompt string `json:"prompt"`
+
+	// KeepAlive controls how long the model will stay loaded in memory following
+	// this request.
+	KeepAlive *Duration `json:"keep_alive,omitempty"`
+
+	// Options lists model-specific options.
+	Options map[string]interface{} `json:"options"`
+}
+
+// EmbeddingResponse is the response from [Client.Embeddings].
 type EmbeddingResponse struct {
 	Embedding []float64 `json:"embedding"`
 }
 
+// CreateRequest is the request passed to [Client.Create].
 type CreateRequest struct {
 	Model     string `json:"model"`
 	Path      string `json:"path"`
 	Modelfile string `json:"modelfile"`
 	Stream    *bool  `json:"stream,omitempty"`
+	Quantize  string `json:"quantize,omitempty"`
 
 	// Name is deprecated, see Model
 	Name string `json:"name"`
+
+	// Quantization is deprecated, see Quantize
+	Quantization string `json:"quantization,omitempty"`
 }
 
+// DeleteRequest is the request passed to [Client.Delete].
 type DeleteRequest struct {
 	Model string `json:"model"`
 
@@ -156,10 +307,14 @@ type DeleteRequest struct {
 	Name string `json:"name"`
 }
 
+// ShowRequest is the request passed to [Client.Show].
 type ShowRequest struct {
-	Model    string `json:"model"`
-	System   string `json:"system"`
+	Model  string `json:"model"`
+	System string `json:"system"`
+
+	// Template is deprecated
 	Template string `json:"template"`
+	Verbose  bool   `json:"verbose"`
 
 	Options map[string]interface{} `json:"options"`
 
@@ -167,21 +322,27 @@ type ShowRequest struct {
 	Name string `json:"name"`
 }
 
+// ShowResponse is the response returned from [Client.Show].
 type ShowResponse struct {
-	License    string       `json:"license,omitempty"`
-	Modelfile  string       `json:"modelfile,omitempty"`
-	Parameters string       `json:"parameters,omitempty"`
-	Template   string       `json:"template,omitempty"`
-	System     string       `json:"system,omitempty"`
-	Details    ModelDetails `json:"details,omitempty"`
-	Messages   []Message    `json:"messages,omitempty"`
+	License       string         `json:"license,omitempty"`
+	Modelfile     string         `json:"modelfile,omitempty"`
+	Parameters    string         `json:"parameters,omitempty"`
+	Template      string         `json:"template,omitempty"`
+	System        string         `json:"system,omitempty"`
+	Details       ModelDetails   `json:"details,omitempty"`
+	Messages      []Message      `json:"messages,omitempty"`
+	ModelInfo     map[string]any `json:"model_info,omitempty"`
+	ProjectorInfo map[string]any `json:"projector_info,omitempty"`
+	ModifiedAt    time.Time      `json:"modified_at,omitempty"`
 }
 
+// CopyRequest is the request passed to [Client.Copy].
 type CopyRequest struct {
 	Source      string `json:"source"`
 	Destination string `json:"destination"`
 }
 
+// PullRequest is the request passed to [Client.Pull].
 type PullRequest struct {
 	Model    string `json:"model"`
 	Insecure bool   `json:"insecure,omitempty"`
@@ -193,6 +354,8 @@ type PullRequest struct {
 	Name string `json:"name"`
 }
 
+// ProgressResponse is the response passed to progress functions like
+// [PullProgressFunc] and [PushProgressFunc].
 type ProgressResponse struct {
 	Status    string `json:"status"`
 	Digest    string `json:"digest,omitempty"`
@@ -200,6 +363,7 @@ type ProgressResponse struct {
 	Completed int64  `json:"completed,omitempty"`
 }
 
+// PushRequest is the request passed to [Client.Push].
 type PushRequest struct {
 	Model    string `json:"model"`
 	Insecure bool   `json:"insecure,omitempty"`
@@ -211,11 +375,18 @@ type PushRequest struct {
 	Name string `json:"name"`
 }
 
+// ListResponse is the response from [Client.List].
 type ListResponse struct {
-	Models []ModelResponse `json:"models"`
+	Models []ListModelResponse `json:"models"`
 }
 
-type ModelResponse struct {
+// ProcessResponse is the response from [Client.Process].
+type ProcessResponse struct {
+	Models []ProcessModelResponse `json:"models"`
+}
+
+// ListModelResponse is a single model description in [ListResponse].
+type ListModelResponse struct {
 	Name       string       `json:"name"`
 	Model      string       `json:"model"`
 	ModifiedAt time.Time    `json:"modified_at"`
@@ -224,21 +395,53 @@ type ModelResponse struct {
 	Details    ModelDetails `json:"details,omitempty"`
 }
 
+// ProcessModelResponse is a single model description in [ProcessResponse].
+type ProcessModelResponse struct {
+	Name      string       `json:"name"`
+	Model     string       `json:"model"`
+	Size      int64        `json:"size"`
+	Digest    string       `json:"digest"`
+	Details   ModelDetails `json:"details,omitempty"`
+	ExpiresAt time.Time    `json:"expires_at"`
+	SizeVRAM  int64        `json:"size_vram"`
+}
+
+type RetrieveModelResponse struct {
+	Id      string `json:"id"`
+	Object  string `json:"object"`
+	Created int64  `json:"created"`
+	OwnedBy string `json:"owned_by"`
+}
+
 type TokenResponse struct {
 	Token string `json:"token"`
 }
 
+// GenerateResponse is the response passed into [GenerateResponseFunc].
 type GenerateResponse struct {
-	Model     string    `json:"model"`
-	CreatedAt time.Time `json:"created_at"`
-	Response  string    `json:"response"`
+	// Model is the model name that generated the response.
+	Model string `json:"model"`
 
-	Done    bool  `json:"done"`
+	// CreatedAt is the timestamp of the response.
+	CreatedAt time.Time `json:"created_at"`
+
+	// Response is the textual response itself.
+	Response string `json:"response"`
+
+	// Done specifies if the response is complete.
+	Done bool `json:"done"`
+
+	// DoneReason is the reason the model stopped generating text.
+	DoneReason string `json:"done_reason,omitempty"`
+
+	// Context is an encoding of the conversation used in this response; this
+	// can be sent in the next request to keep a conversational memory.
 	Context []int `json:"context,omitempty"`
 
 	Metrics
 }
 
+// ModelDetails provides details about a model.
 type ModelDetails struct {
 	ParentModel       string   `json:"parent_model"`
 	Format            string   `json:"format"`
@@ -276,8 +479,6 @@ func (m *Metrics) Summary() {
 	}
 }
 
-var ErrInvalidOpts = fmt.Errorf("invalid options")
-
 func (opts *Options) FromMap(m map[string]interface{}) error {
 	valueOpts := reflect.ValueOf(opts).Elem() // names of the fields in the options struct
 	typeOpts := reflect.TypeOf(opts).Elem()   // types of the fields in the options struct
@@ -291,81 +492,94 @@ func (opts *Options) FromMap(m map[string]interface{}) error {
 		}
 	}
 
-	invalidOpts := []string{}
 	for key, val := range m {
-		if opt, ok := jsonOpts[key]; ok {
-			field := valueOpts.FieldByName(opt.Name)
-			if field.IsValid() && field.CanSet() {
-				if val == nil {
-					continue
-				}
+		opt, ok := jsonOpts[key]
+		if !ok {
+			slog.Warn("invalid option provided", "option", opt.Name)
+			continue
+		}
 
-				switch field.Kind() {
-				case reflect.Int:
-					switch t := val.(type) {
-					case int64:
-						field.SetInt(t)
-					case float64:
-						// when JSON unmarshals numbers, it uses float64, not int
-						field.SetInt(int64(t))
-					default:
-						return fmt.Errorf("option %q must be of type integer", key)
+		field := valueOpts.FieldByName(opt.Name)
+		if field.IsValid() && field.CanSet() {
+			if val == nil {
+				continue
+			}
+
+			switch field.Kind() {
+			case reflect.Int:
+				switch t := val.(type) {
+				case int64:
+					field.SetInt(t)
+				case float64:
+					// when JSON unmarshals numbers, it uses float64, not int
+					field.SetInt(int64(t))
+				default:
+					return fmt.Errorf("option %q must be of type integer", key)
+				}
+			case reflect.Bool:
+				val, ok := val.(bool)
+				if !ok {
+					return fmt.Errorf("option %q must be of type boolean", key)
+				}
+				field.SetBool(val)
+			case reflect.Float32:
+				// JSON unmarshals to float64
+				val, ok := val.(float64)
+				if !ok {
+					return fmt.Errorf("option %q must be of type float32", key)
+				}
+				field.SetFloat(val)
+			case reflect.String:
+				val, ok := val.(string)
+				if !ok {
+					return fmt.Errorf("option %q must be of type string", key)
+				}
+				field.SetString(val)
+			case reflect.Slice:
+				// JSON unmarshals to []interface{}, not []string
+				val, ok := val.([]interface{})
+				if !ok {
+					return fmt.Errorf("option %q must be of type array", key)
+				}
+				// convert []interface{} to []string
+				slice := make([]string, len(val))
+				for i, item := range val {
+					str, ok := item.(string)
+					if !ok {
+						return fmt.Errorf("option %q must be of an array of strings", key)
 					}
-				case reflect.Bool:
+					slice[i] = str
+				}
+				field.Set(reflect.ValueOf(slice))
+			case reflect.Pointer:
+				var b bool
+				if field.Type() == reflect.TypeOf(&b) {
 					val, ok := val.(bool)
 					if !ok {
 						return fmt.Errorf("option %q must be of type boolean", key)
 					}
-					field.SetBool(val)
-				case reflect.Float32:
-					// JSON unmarshals to float64
-					val, ok := val.(float64)
-					if !ok {
-						return fmt.Errorf("option %q must be of type float32", key)
-					}
-					field.SetFloat(val)
-				case reflect.String:
-					val, ok := val.(string)
-					if !ok {
-						return fmt.Errorf("option %q must be of type string", key)
-					}
-					field.SetString(val)
-				case reflect.Slice:
-					// JSON unmarshals to []interface{}, not []string
-					val, ok := val.([]interface{})
-					if !ok {
-						return fmt.Errorf("option %q must be of type array", key)
-					}
-					// convert []interface{} to []string
-					slice := make([]string, len(val))
-					for i, item := range val {
-						str, ok := item.(string)
-						if !ok {
-							return fmt.Errorf("option %q must be of an array of strings", key)
-						}
-						slice[i] = str
-					}
-					field.Set(reflect.ValueOf(slice))
-				default:
-					return fmt.Errorf("unknown type loading config params: %v", field.Kind())
+					field.Set(reflect.ValueOf(&val))
+				} else {
+					return fmt.Errorf("unknown type loading config params: %v %v", field.Kind(), field.Type())
 				}
+			default:
+				return fmt.Errorf("unknown type loading config params: %v", field.Kind())
 			}
-		} else {
-			invalidOpts = append(invalidOpts, key)
 		}
 	}
 
-	if len(invalidOpts) > 0 {
-		return fmt.Errorf("%w: %v", ErrInvalidOpts, strings.Join(invalidOpts, ", "))
-	}
 	return nil
 }
 
+// DefaultOptions is the default set of options for [GenerateRequest]; these
+// values are used unless the user specifies other values explicitly.
 func DefaultOptions() Options {
 	return Options{
 		// options set on request to runner
-		NumPredict:       -1,
-		NumKeep:          0,
+		NumPredict: -1,
+
+		// set a minimal num_keep to avoid issues on context shifts
+		NumKeep:          4,
 		Temperature:      0.8,
 		TopK:             40,
 		TopP:             0.9,
@@ -383,19 +597,15 @@ func DefaultOptions() Options {
 
 		Runner: Runner{
 			// options set when the model is loaded
-			NumCtx:             2048,
-			RopeFrequencyBase:  10000.0,
-			RopeFrequencyScale: 1.0,
-			NumBatch:           512,
-			NumGPU:             -1, // -1 here indicates that NumGPU should be set dynamically
-			NumGQA:             1,
-			NumThread:          0, // let the runtime decide
-			LowVRAM:            false,
-			F16KV:              true,
-			UseMLock:           false,
-			UseMMap:            true,
-			UseNUMA:            false,
-			EmbeddingOnly:      true,
+			NumCtx:    2048,
+			NumBatch:  512,
+			NumGPU:    -1, // -1 here indicates that NumGPU should be set dynamically
+			NumThread: 0,  // let the runtime decide
+			LowVRAM:   false,
+			F16KV:     true,
+			UseMLock:  false,
+			UseMMap:   nil,
+			UseNUMA:   false,
 		},
 	}
 }
@@ -404,6 +614,13 @@ type Duration struct {
 	time.Duration
 }
 
+func (d Duration) MarshalJSON() ([]byte, error) {
+	if d.Duration < 0 {
+		return []byte("-1"), nil
+	}
+	return []byte("\"" + d.Duration.String() + "\""), nil
+}
+
 func (d *Duration) UnmarshalJSON(b []byte) (err error) {
 	var v any
 	if err := json.Unmarshal(b, &v); err != nil {
@@ -417,7 +634,7 @@ func (d *Duration) UnmarshalJSON(b []byte) (err error) {
 		if t < 0 {
 			d.Duration = time.Duration(math.MaxInt64)
 		} else {
-			d.Duration = time.Duration(t * float64(time.Second))
+			d.Duration = time.Duration(int(t) * int(time.Second))
 		}
 	case string:
 		d.Duration, err = time.ParseDuration(t)
@@ -427,6 +644,8 @@ func (d *Duration) UnmarshalJSON(b []byte) (err error) {
 		if d.Duration < 0 {
 			d.Duration = time.Duration(math.MaxInt64)
 		}
+	default:
+		return fmt.Errorf("Unsupported type: '%s'", reflect.TypeOf(v))
 	}
 
 	return nil
@@ -482,6 +701,17 @@ func FormatParams(params map[string][]string) (map[string]interface{}, error) {
 				case reflect.Slice:
 					// TODO: only string slices are supported right now
 					out[key] = vals
+				case reflect.Pointer:
+					var b bool
+					if field.Type() == reflect.TypeOf(&b) {
+						boolVal, err := strconv.ParseBool(vals[0])
+						if err != nil {
+							return nil, fmt.Errorf("invalid bool value %s", vals)
+						}
+						out[key] = &boolVal
+					} else {
+						return nil, fmt.Errorf("unknown type %s for %s", field.Kind(), key)
+					}
 				default:
 					return nil, fmt.Errorf("unknown type %s for %s", field.Kind(), key)
 				}

api/types_test.go

@@ -0,0 +1,233 @@
+package api
+
+import (
+	"encoding/json"
+	"fmt"
+	"math"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestKeepAliveParsingFromJSON(t *testing.T) {
+	tests := []struct {
+		name string
+		req  string
+		exp  *Duration
+	}{
+		{
+			name: "Positive Integer",
+			req:  `{ "keep_alive": 42 }`,
+			exp:  &Duration{42 * time.Second},
+		},
+		{
+			name: "Positive Float",
+			req:  `{ "keep_alive": 42.5 }`,
+			exp:  &Duration{42 * time.Second},
+		},
+		{
+			name: "Positive Integer String",
+			req:  `{ "keep_alive": "42m" }`,
+			exp:  &Duration{42 * time.Minute},
+		},
+		{
+			name: "Negative Integer",
+			req:  `{ "keep_alive": -1 }`,
+			exp:  &Duration{math.MaxInt64},
+		},
+		{
+			name: "Negative Float",
+			req:  `{ "keep_alive": -3.14 }`,
+			exp:  &Duration{math.MaxInt64},
+		},
+		{
+			name: "Negative Integer String",
+			req:  `{ "keep_alive": "-1m" }`,
+			exp:  &Duration{math.MaxInt64},
+		},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			var dec ChatRequest
+			err := json.Unmarshal([]byte(test.req), &dec)
+			require.NoError(t, err)
+
+			assert.Equal(t, test.exp, dec.KeepAlive)
+		})
+	}
+}
+
+func TestDurationMarshalUnmarshal(t *testing.T) {
+	tests := []struct {
+		name     string
+		input    time.Duration
+		expected time.Duration
+	}{
+		{
+			"negative duration",
+			time.Duration(-1),
+			time.Duration(math.MaxInt64),
+		},
+		{
+			"positive duration",
+			42 * time.Second,
+			42 * time.Second,
+		},
+		{
+			"another positive duration",
+			42 * time.Minute,
+			42 * time.Minute,
+		},
+		{
+			"zero duration",
+			time.Duration(0),
+			time.Duration(0),
+		},
+		{
+			"max duration",
+			time.Duration(math.MaxInt64),
+			time.Duration(math.MaxInt64),
+		},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			b, err := json.Marshal(Duration{test.input})
+			require.NoError(t, err)
+
+			var d Duration
+			err = json.Unmarshal(b, &d)
+			require.NoError(t, err)
+
+			assert.Equal(t, test.expected, d.Duration, "input %v, marshalled %v, got %v", test.input, string(b), d.Duration)
+		})
+	}
+}
+
+func TestUseMmapParsingFromJSON(t *testing.T) {
+	tr := true
+	fa := false
+	tests := []struct {
+		name string
+		req  string
+		exp  *bool
+	}{
+		{
+			name: "Undefined",
+			req:  `{ }`,
+			exp:  nil,
+		},
+		{
+			name: "True",
+			req:  `{ "use_mmap": true }`,
+			exp:  &tr,
+		},
+		{
+			name: "False",
+			req:  `{ "use_mmap": false }`,
+			exp:  &fa,
+		},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			var oMap map[string]interface{}
+			err := json.Unmarshal([]byte(test.req), &oMap)
+			require.NoError(t, err)
+			opts := DefaultOptions()
+			err = opts.FromMap(oMap)
+			require.NoError(t, err)
+			assert.Equal(t, test.exp, opts.UseMMap)
+		})
+	}
+}
+
+func TestUseMmapFormatParams(t *testing.T) {
+	tr := true
+	fa := false
+	tests := []struct {
+		name string
+		req  map[string][]string
+		exp  *bool
+		err  error
+	}{
+		{
+			name: "True",
+			req: map[string][]string{
+				"use_mmap": {"true"},
+			},
+			exp: &tr,
+			err: nil,
+		},
+		{
+			name: "False",
+			req: map[string][]string{
+				"use_mmap": {"false"},
+			},
+			exp: &fa,
+			err: nil,
+		},
+		{
+			name: "Numeric True",
+			req: map[string][]string{
+				"use_mmap": {"1"},
+			},
+			exp: &tr,
+			err: nil,
+		},
+		{
+			name: "Numeric False",
+			req: map[string][]string{
+				"use_mmap": {"0"},
+			},
+			exp: &fa,
+			err: nil,
+		},
+		{
+			name: "invalid string",
+			req: map[string][]string{
+				"use_mmap": {"foo"},
+			},
+			exp: nil,
+			err: fmt.Errorf("invalid bool value [foo]"),
+		},
+	}
+
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			resp, err := FormatParams(test.req)
+			require.Equal(t, test.err, err)
+			respVal, ok := resp["use_mmap"]
+			if test.exp != nil {
+				assert.True(t, ok, "resp: %v", resp)
+				assert.Equal(t, *test.exp, *respVal.(*bool))
+			}
+		})
+	}
+}
+
+func TestMessage_UnmarshalJSON(t *testing.T) {
+	tests := []struct {
+		input    string
+		expected string
+	}{
+		{`{"role": "USER", "content": "Hello!"}`, "user"},
+		{`{"role": "System", "content": "Initialization complete."}`, "system"},
+		{`{"role": "assistant", "content": "How can I help you?"}`, "assistant"},
+		{`{"role": "TOOl", "content": "Access granted."}`, "tool"},
+	}
+
+	for _, test := range tests {
+		var msg Message
+		if err := json.Unmarshal([]byte(test.input), &msg); err != nil {
+			t.Errorf("Unexpected error: %v", err)
+		}
+
+		if msg.Role != test.expected {
+			t.Errorf("role not lowercased: got %v, expected %v", msg.Role, test.expected)
+		}
+	}
+}

app/lifecycle/lifecycle.go

@@ -9,8 +9,8 @@ import (
 	"os/signal"
 	"syscall"
 
-	"github.com/jmorganca/ollama/app/store"
-	"github.com/jmorganca/ollama/app/tray"
+	"github.com/ollama/ollama/app/store"
+	"github.com/ollama/ollama/app/tray"
 )
 
 func Run() {

app/lifecycle/logging.go

@@ -5,12 +5,16 @@ import (
 	"log/slog"
 	"os"
 	"path/filepath"
+	"strconv"
+	"strings"
+
+	"github.com/ollama/ollama/envconfig"
 )
 
 func InitLogging() {
 	level := slog.LevelInfo
 
-	if debug := os.Getenv("OLLAMA_DEBUG"); debug != "" {
+	if envconfig.Debug {
 		level = slog.LevelDebug
 	}
 
@@ -22,6 +26,7 @@ func InitLogging() {
 		logFile = os.Stderr
 		// TODO - write one-line to the app.log file saying we're running in console mode to help avoid confusion
 	} else {
+		rotateLogs(AppLogFile)
 		logFile, err = os.OpenFile(AppLogFile, os.O_APPEND|os.O_WRONLY|os.O_CREATE, 0755)
 		if err != nil {
 			slog.Error(fmt.Sprintf("failed to create server log %v", err))
@@ -44,3 +49,32 @@ func InitLogging() {
 
 	slog.Info("ollama app started")
 }
+
+func rotateLogs(logFile string) {
+	if _, err := os.Stat(logFile); os.IsNotExist(err) {
+		return
+	}
+	index := strings.LastIndex(logFile, ".")
+	pre := logFile[:index]
+	post := "." + logFile[index+1:]
+	for i := LogRotationCount; i > 0; i-- {
+		older := pre + "-" + strconv.Itoa(i) + post
+		newer := pre + "-" + strconv.Itoa(i-1) + post
+		if i == 1 {
+			newer = pre + post
+		}
+		if _, err := os.Stat(newer); err == nil {
+			if _, err := os.Stat(older); err == nil {
+				err := os.Remove(older)
+				if err != nil {
+					slog.Warn("Failed to remove older log", "older", older, "error", err)
+					continue
+				}
+			}
+			err := os.Rename(newer, older)
+			if err != nil {
+				slog.Warn("Failed to rotate log", "older", older, "newer", newer, "error", err)
+			}
+		}
+	}
+}

app/lifecycle/logging_test.go

@@ -0,0 +1,44 @@
+package lifecycle
+
+import (
+	"os"
+	"path/filepath"
+	"strconv"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestRotateLogs(t *testing.T) {
+	logDir := t.TempDir()
+	logFile := filepath.Join(logDir, "testlog.log")
+
+	// No log exists
+	rotateLogs(logFile)
+
+	require.NoError(t, os.WriteFile(logFile, []byte("1"), 0644))
+	assert.FileExists(t, logFile)
+	// First rotation
+	rotateLogs(logFile)
+	assert.FileExists(t, filepath.Join(logDir, "testlog-1.log"))
+	assert.NoFileExists(t, filepath.Join(logDir, "testlog-2.log"))
+	assert.NoFileExists(t, logFile)
+
+	// Should be a no-op without a new log
+	rotateLogs(logFile)
+	assert.FileExists(t, filepath.Join(logDir, "testlog-1.log"))
+	assert.NoFileExists(t, filepath.Join(logDir, "testlog-2.log"))
+	assert.NoFileExists(t, logFile)
+
+	for i := 2; i <= LogRotationCount+1; i++ {
+		require.NoError(t, os.WriteFile(logFile, []byte(strconv.Itoa(i)), 0644))
+		assert.FileExists(t, logFile)
+		rotateLogs(logFile)
+		assert.NoFileExists(t, logFile)
+		for j := 1; j < i; j++ {
+			assert.FileExists(t, filepath.Join(logDir, "testlog-"+strconv.Itoa(j)+".log"))
+		}
+		assert.NoFileExists(t, filepath.Join(logDir, "testlog-"+strconv.Itoa(i+1)+".log"))
+	}
+}

app/lifecycle/paths.go

@@ -16,11 +16,12 @@ var (
 	AppDir     = "/opt/Ollama"
 	AppDataDir = "/opt/Ollama"
 	// TODO - should there be a distinct log dir?
-	UpdateStageDir = "/tmp"
-	AppLogFile     = "/tmp/ollama_app.log"
-	ServerLogFile  = "/tmp/ollama.log"
-	UpgradeLogFile = "/tmp/ollama_update.log"
-	Installer      = "OllamaSetup.exe"
+	UpdateStageDir   = "/tmp"
+	AppLogFile       = "/tmp/ollama_app.log"
+	ServerLogFile    = "/tmp/ollama.log"
+	UpgradeLogFile   = "/tmp/ollama_update.log"
+	Installer        = "OllamaSetup.exe"
+	LogRotationCount = 5
 )
 
 func init() {
@@ -69,7 +70,6 @@ func init() {
 				slog.Error(fmt.Sprintf("create ollama dir %s: %v", AppDataDir, err))
 			}
 		}
-
 	} else if runtime.GOOS == "darwin" {
 		// TODO
 		AppName += ".app"

app/lifecycle/server.go

@@ -11,11 +11,11 @@ import (
 	"path/filepath"
 	"time"
 
-	"github.com/jmorganca/ollama/api"
+	"github.com/ollama/ollama/api"
 )
 
 func getCLIFullPath(command string) string {
-	cmdPath := ""
+	var cmdPath string
 	appExe, err := os.Executable()
 	if err == nil {
 		cmdPath = filepath.Join(filepath.Dir(appExe), command)
@@ -43,37 +43,35 @@ func getCLIFullPath(command string) string {
 	return command
 }
 
-func SpawnServer(ctx context.Context, command string) (chan int, error) {
-	done := make(chan int)
-
-	logDir := filepath.Dir(ServerLogFile)
-	_, err := os.Stat(logDir)
-	if errors.Is(err, os.ErrNotExist) {
-		if err := os.MkdirAll(logDir, 0o755); err != nil {
-			return done, fmt.Errorf("create ollama server log dir %s: %v", logDir, err)
-		}
-	}
-
+func start(ctx context.Context, command string) (*exec.Cmd, error) {
 	cmd := getCmd(ctx, getCLIFullPath(command))
-	// send stdout and stderr to a file
 	stdout, err := cmd.StdoutPipe()
 	if err != nil {
-		return done, fmt.Errorf("failed to spawn server stdout pipe %s", err)
+		return nil, fmt.Errorf("failed to spawn server stdout pipe: %w", err)
 	}
 	stderr, err := cmd.StderrPipe()
 	if err != nil {
-		return done, fmt.Errorf("failed to spawn server stderr pipe %s", err)
-	}
-	stdin, err := cmd.StdinPipe()
-	if err != nil {
-		return done, fmt.Errorf("failed to spawn server stdin pipe %s", err)
+		return nil, fmt.Errorf("failed to spawn server stderr pipe: %w", err)
 	}
 
-	// TODO - rotation
+	rotateLogs(ServerLogFile)
 	logFile, err := os.OpenFile(ServerLogFile, os.O_APPEND|os.O_WRONLY|os.O_CREATE, 0755)
 	if err != nil {
-		return done, fmt.Errorf("failed to create server log %w", err)
+		return nil, fmt.Errorf("failed to create server log: %w", err)
 	}
+
+	logDir := filepath.Dir(ServerLogFile)
+	_, err = os.Stat(logDir)
+	if err != nil {
+		if !errors.Is(err, os.ErrNotExist) {
+			return nil, fmt.Errorf("stat ollama server log dir %s: %v", logDir, err)
+		}
+
+		if err := os.MkdirAll(logDir, 0o755); err != nil {
+			return nil, fmt.Errorf("create ollama server log dir %s: %v", logDir, err)
+		}
+	}
+
 	go func() {
 		defer logFile.Close()
 		io.Copy(logFile, stdout) //nolint:errcheck
@@ -83,21 +81,67 @@ func SpawnServer(ctx context.Context, command string) (chan int, error) {
 		io.Copy(logFile, stderr) //nolint:errcheck
 	}()
 
+	// Re-wire context done behavior to attempt a graceful shutdown of the server
+	cmd.Cancel = func() error {
+		if cmd.Process != nil {
+			err := terminate(cmd)
+			if err != nil {
+				slog.Warn("error trying to gracefully terminate server", "err", err)
+				return cmd.Process.Kill()
+			}
+
+			tick := time.NewTicker(10 * time.Millisecond)
+			defer tick.Stop()
+
+			for {
+				select {
+				case <-tick.C:
+					exited, err := isProcessExited(cmd.Process.Pid)
+					if err != nil {
+						return err
+					}
+
+					if exited {
+						return nil
+					}
+				case <-time.After(5 * time.Second):
+					slog.Warn("graceful server shutdown timeout, killing", "pid", cmd.Process.Pid)
+					return cmd.Process.Kill()
+				}
+			}
+		}
+		return nil
+	}
+
 	// run the command and wait for it to finish
 	if err := cmd.Start(); err != nil {
-		return done, fmt.Errorf("failed to start server %w", err)
+		return nil, fmt.Errorf("failed to start server %w", err)
 	}
 	if cmd.Process != nil {
 		slog.Info(fmt.Sprintf("started ollama server with pid %d", cmd.Process.Pid))
 	}
 	slog.Info(fmt.Sprintf("ollama server logs %s", ServerLogFile))
 
+	return cmd, nil
+}
+
+func SpawnServer(ctx context.Context, command string) (chan int, error) {
+	done := make(chan int)
+
 	go func() {
 		// Keep the server running unless we're shuttind down the app
 		crashCount := 0
 		for {
+			slog.Info("starting server...")
+			cmd, err := start(ctx, command)
+			if err != nil {
+				crashCount++
+				slog.Error(fmt.Sprintf("failed to start server %s", err))
+				time.Sleep(500 * time.Millisecond * time.Duration(crashCount))
+				continue
+			}
+
 			cmd.Wait() //nolint:errcheck
-			stdin.Close()
 			var code int
 			if cmd.ProcessState != nil {
 				code = cmd.ProcessState.ExitCode()
@@ -105,21 +149,18 @@ func SpawnServer(ctx context.Context, command string) (chan int, error) {
 
 			select {
 			case <-ctx.Done():
-				slog.Debug(fmt.Sprintf("server shutdown with exit code %d", code))
+				slog.Info(fmt.Sprintf("server shutdown with exit code %d", code))
 				done <- code
 				return
 			default:
 				crashCount++
 				slog.Warn(fmt.Sprintf("server crash %d - exit code %d - respawning", crashCount, code))
-				time.Sleep(500 * time.Millisecond)
-				if err := cmd.Start(); err != nil {
-					slog.Error(fmt.Sprintf("failed to restart server %s", err))
-					// Keep trying, but back off if we keep failing
-					time.Sleep(time.Duration(crashCount) * time.Second)
-				}
+				time.Sleep(500 * time.Millisecond * time.Duration(crashCount))
+				break
 			}
 		}
 	}()
+
 	return done, nil
 }
 

app/lifecycle/server_unix.go

@@ -4,9 +4,35 @@ package lifecycle
 
 import (
 	"context"
+	"errors"
+	"fmt"
+	"os"
 	"os/exec"
+	"syscall"
 )
 
 func getCmd(ctx context.Context, cmd string) *exec.Cmd {
 	return exec.CommandContext(ctx, cmd, "serve")
 }
+
+func terminate(cmd *exec.Cmd) error {
+	return cmd.Process.Signal(os.Interrupt)
+}
+
+func isProcessExited(pid int) (bool, error) {
+	proc, err := os.FindProcess(pid)
+	if err != nil {
+		return false, fmt.Errorf("failed to find process: %v", err)
+	}
+
+	err = proc.Signal(syscall.Signal(0))
+	if err != nil {
+		if errors.Is(err, os.ErrProcessDone) || errors.Is(err, syscall.ESRCH) {
+			return true, nil
+		}
+
+		return false, fmt.Errorf("error signaling process: %v", err)
+	}
+
+	return false, nil
+}

app/lifecycle/server_windows.go

@@ -2,12 +2,90 @@ package lifecycle
 
 import (
 	"context"
+	"fmt"
 	"os/exec"
 	"syscall"
+
+	"golang.org/x/sys/windows"
 )
 
 func getCmd(ctx context.Context, exePath string) *exec.Cmd {
 	cmd := exec.CommandContext(ctx, exePath, "serve")
-	cmd.SysProcAttr = &syscall.SysProcAttr{HideWindow: true, CreationFlags: 0x08000000}
+	cmd.SysProcAttr = &syscall.SysProcAttr{
+		HideWindow:    true,
+		CreationFlags: windows.CREATE_NEW_PROCESS_GROUP,
+	}
+
 	return cmd
 }
+
+func terminate(cmd *exec.Cmd) error {
+	dll, err := windows.LoadDLL("kernel32.dll")
+	if err != nil {
+		return err
+	}
+	//nolint:errcheck
+	defer dll.Release()
+
+	pid := cmd.Process.Pid
+
+	f, err := dll.FindProc("AttachConsole")
+	if err != nil {
+		return err
+	}
+
+	r1, _, err := f.Call(uintptr(pid))
+	if r1 == 0 && err != syscall.ERROR_ACCESS_DENIED {
+		return err
+	}
+
+	f, err = dll.FindProc("SetConsoleCtrlHandler")
+	if err != nil {
+		return err
+	}
+
+	r1, _, err = f.Call(0, 1)
+	if r1 == 0 {
+		return err
+	}
+
+	f, err = dll.FindProc("GenerateConsoleCtrlEvent")
+	if err != nil {
+		return err
+	}
+
+	r1, _, err = f.Call(windows.CTRL_BREAK_EVENT, uintptr(pid))
+	if r1 == 0 {
+		return err
+	}
+
+	r1, _, err = f.Call(windows.CTRL_C_EVENT, uintptr(pid))
+	if r1 == 0 {
+		return err
+	}
+
+	return nil
+}
+
+const STILL_ACTIVE = 259
+
+func isProcessExited(pid int) (bool, error) {
+	hProcess, err := windows.OpenProcess(windows.PROCESS_QUERY_INFORMATION, false, uint32(pid))
+	if err != nil {
+		return false, fmt.Errorf("failed to open process: %v", err)
+	}
+	//nolint:errcheck
+	defer windows.CloseHandle(hProcess)
+
+	var exitCode uint32
+	err = windows.GetExitCodeProcess(hProcess, &exitCode)
+	if err != nil {
+		return false, fmt.Errorf("failed to get exit code: %v", err)
+	}
+
+	if exitCode == STILL_ACTIVE {
+		return false, nil
+	}
+
+	return true, nil
+}

app/lifecycle/updater.go

@@ -18,8 +18,8 @@ import (
 	"strings"
 	"time"
 
-	"github.com/jmorganca/ollama/auth"
-	"github.com/jmorganca/ollama/version"
+	"github.com/ollama/ollama/auth"
+	"github.com/ollama/ollama/version"
 )
 
 var (
@@ -34,20 +34,6 @@ type UpdateResponse struct {
 	UpdateVersion string `json:"version"`
 }
 
-func getClient(req *http.Request) http.Client {
-	proxyURL, err := http.ProxyFromEnvironment(req)
-	if err != nil {
-		slog.Warn(fmt.Sprintf("failed to handle proxy: %s", err))
-		return http.Client{}
-	}
-
-	return http.Client{
-		Transport: &http.Transport{
-			Proxy: http.ProxyURL(proxyURL),
-		},
-	}
-}
-
 func IsNewReleaseAvailable(ctx context.Context) (bool, UpdateResponse) {
 	var updateResp UpdateResponse
 
@@ -83,17 +69,16 @@ func IsNewReleaseAvailable(ctx context.Context) (bool, UpdateResponse) {
 	}
 	req.Header.Set("Authorization", signature)
 	req.Header.Set("User-Agent", fmt.Sprintf("ollama/%s (%s %s) Go/%s", version.Version, runtime.GOARCH, runtime.GOOS, runtime.Version()))
-	client := getClient(req)
 
 	slog.Debug("checking for available update", "requestURL", requestURL)
-	resp, err := client.Do(req)
+	resp, err := http.DefaultClient.Do(req)
 	if err != nil {
 		slog.Warn(fmt.Sprintf("failed to check for update: %s", err))
 		return false, updateResp
 	}
 	defer resp.Body.Close()
 
-	if resp.StatusCode == 204 {
+	if resp.StatusCode == http.StatusNoContent {
 		slog.Debug("check update response 204 (current version is up to date)")
 		return false, updateResp
 	}
@@ -101,6 +86,11 @@ func IsNewReleaseAvailable(ctx context.Context) (bool, UpdateResponse) {
 	if err != nil {
 		slog.Warn(fmt.Sprintf("failed to read body response: %s", err))
 	}
+
+	if resp.StatusCode != http.StatusOK {
+		slog.Info(fmt.Sprintf("check update error %d - %.96s", resp.StatusCode, string(body)))
+		return false, updateResp
+	}
 	err = json.Unmarshal(body, &updateResp)
 	if err != nil {
 		slog.Warn(fmt.Sprintf("malformed response checking for update: %s", err))
@@ -119,12 +109,12 @@ func DownloadNewRelease(ctx context.Context, updateResp UpdateResponse) error {
 	if err != nil {
 		return err
 	}
-	client := getClient(req)
-	resp, err := client.Do(req)
+
+	resp, err := http.DefaultClient.Do(req)
 	if err != nil {
 		return fmt.Errorf("error checking update: %w", err)
 	}
-	if resp.StatusCode != 200 {
+	if resp.StatusCode != http.StatusOK {
 		return fmt.Errorf("unexpected status attempting to download update %d", resp.StatusCode)
 	}
 	resp.Body.Close()
@@ -151,7 +141,7 @@ func DownloadNewRelease(ctx context.Context, updateResp UpdateResponse) error {
 	cleanupOldDownloads()
 
 	req.Method = http.MethodGet
-	resp, err = client.Do(req)
+	resp, err = http.DefaultClient.Do(req)
 	if err != nil {
 		return fmt.Errorf("error checking update: %w", err)
 	}

app/lifecycle/updater_windows.go

@@ -31,16 +31,13 @@ func DoUpgrade(cancel context.CancelFunc, done chan int) error {
 		"/LOG=" + filepath.Base(UpgradeLogFile), // Only relative seems reliable, so set pwd
 		"/FORCECLOSEAPPLICATIONS",               // Force close the tray app - might be needed
 	}
-	// When we're not in debug mode, make the upgrade as quiet as possible (no GUI, no prompts)
-	// TODO - temporarily disable since we're pinning in debug mode for the preview
-	// if debug := os.Getenv("OLLAMA_DEBUG"); debug == "" {
+	// make the upgrade as quiet as possible (no GUI, no prompts)
 	installArgs = append(installArgs,
 		"/SP", // Skip the "This will install... Do you wish to continue" prompt
 		"/SUPPRESSMSGBOXES",
 		"/SILENT",
 		"/VERYSILENT",
 	)
-	// }
 
 	// Safeguard in case we have requests in flight that need to drain...
 	slog.Info("Waiting for server to shutdown")

app/main.go

@@ -4,7 +4,7 @@ package main
 // go build -ldflags="-H windowsgui" .
 
 import (
-	"github.com/jmorganca/ollama/app/lifecycle"
+	"github.com/ollama/ollama/app/lifecycle"
 )
 
 func main() {

app/ollama.iss

@@ -28,8 +28,8 @@ AppPublisher={#MyAppPublisher}
 AppPublisherURL={#MyAppURL}
 AppSupportURL={#MyAppURL}
 AppUpdatesURL={#MyAppURL}
-ArchitecturesAllowed=x64
-ArchitecturesInstallIn64BitMode=x64
+ArchitecturesAllowed=x64 arm64
+ArchitecturesInstallIn64BitMode=x64 arm64
 DefaultDirName={localappdata}\Programs\{#MyAppName}
 DefaultGroupName={#MyAppName}
 DisableProgramGroupPage=yes
@@ -49,9 +49,6 @@ SetupLogging=yes
 CloseApplications=yes
 RestartApplications=no
 
-; Make sure they can at least download llama2 as a minimum
-ExtraDiskSpaceRequired=3826806784
-
 ; https://jrsoftware.org/ishelp/index.php?topic=setup_wizardimagefile
 WizardSmallImageFile=.\assets\setup.bmp
 
@@ -91,9 +88,19 @@ DialogFontSize=12
 [Files]
 Source: ".\app.exe"; DestDir: "{app}"; DestName: "{#MyAppExeName}" ; Flags: ignoreversion 64bit
 Source: "..\ollama.exe"; DestDir: "{app}"; Flags: ignoreversion 64bit
-Source: "..\dist\windeps\*.dll"; DestDir: "{app}"; Flags: ignoreversion 64bit
+Source: "..\dist\windows-{#ARCH}\ollama_runners\*"; DestDir: "{app}\ollama_runners"; Flags: ignoreversion 64bit recursesubdirs
 Source: "..\dist\ollama_welcome.ps1"; DestDir: "{app}"; Flags: ignoreversion
 Source: ".\assets\app.ico"; DestDir: "{app}"; Flags: ignoreversion
+#if DirExists("..\dist\windows-amd64\cuda")
+  Source: "..\dist\windows-amd64\cuda\*"; DestDir: "{app}\cuda\"; Flags: ignoreversion recursesubdirs
+#endif
+#if DirExists("..\dist\windows-amd64\oneapi")
+  Source: "..\dist\windows-amd64\oneapi\*"; DestDir: "{app}\oneapi\"; Flags: ignoreversion recursesubdirs
+#endif
+#if DirExists("..\dist\windows-amd64\rocm")
+  Source: "..\dist\windows-amd64\rocm\*"; DestDir: "{app}\rocm\"; Flags: ignoreversion recursesubdirs
+#endif
+
 
 [Icons]
 Name: "{group}\{#MyAppName}"; Filename: "{app}\{#MyAppExeName}"; IconFilename: "{app}\app.ico"
@@ -116,9 +123,14 @@ Filename: "{cmd}"; Parameters: "/c timeout 5"; Flags: runhidden
 Type: filesandordirs; Name: "{%TEMP}\ollama*"
 Type: filesandordirs; Name: "{%LOCALAPPDATA}\Ollama"
 Type: filesandordirs; Name: "{%LOCALAPPDATA}\Programs\Ollama"
-Type: filesandordirs; Name: "{%USERPROFILE}\.ollama"
+Type: filesandordirs; Name: "{%USERPROFILE}\.ollama\models"
+Type: filesandordirs; Name: "{%USERPROFILE}\.ollama\history"
 ; NOTE: if the user has a custom OLLAMA_MODELS it will be preserved
 
+[InstallDelete]
+Type: filesandordirs; Name: "{%TEMP}\ollama*"
+Type: filesandordirs; Name: "{%LOCALAPPDATA}\Programs\Ollama"
+
 [Messages]
 WizardReady=Ollama Windows Preview
 ReadyLabel1=%nLet's get you up and running with your own large language models.
@@ -126,7 +138,7 @@ SetupAppRunningError=Another Ollama installer is running.%n%nPlease cancel or fi
 
 
 ;FinishedHeadingLabel=Run your first model
-;FinishedLabel=%nRun this command in a PowerShell or cmd terminal.%n%n%n    ollama run llama2
+;FinishedLabel=%nRun this command in a PowerShell or cmd terminal.%n%n%n    ollama run llama3
 ;ClickFinish=%n
 
 [Registry]

app/ollama_welcome.ps1

@@ -4,5 +4,5 @@ write-host "Welcome to Ollama!"
 write-host ""
 write-host "Run your first model:"
 write-host ""
-write-host "`tollama run llama2"
+write-host "`tollama run llama3"
 write-host ""

app/store/store.go

@@ -29,7 +29,6 @@ func GetID() string {
 		initStore()
 	}
 	return store.ID
-
 }
 
 func GetFirstTimeRun() bool {

app/tray/tray.go

@@ -4,8 +4,8 @@ import (
 	"fmt"
 	"runtime"
 
-	"github.com/jmorganca/ollama/app/assets"
-	"github.com/jmorganca/ollama/app/tray/commontray"
+	"github.com/ollama/ollama/app/assets"
+	"github.com/ollama/ollama/app/tray/commontray"
 )
 
 func NewTray() (commontray.OllamaTray, error) {
@@ -24,10 +24,5 @@ func NewTray() (commontray.OllamaTray, error) {
 		return nil, fmt.Errorf("failed to load icon %s: %w", iconName, err)
 	}
 
-	tray, err := InitPlatformTray(icon, updateIcon)
-	if err != nil {
-		return nil, err
-	}
-
-	return tray, nil
+	return InitPlatformTray(icon, updateIcon)
 }

app/tray/tray_nonwindows.go

@@ -5,7 +5,7 @@ package tray
 import (
 	"fmt"
 
-	"github.com/jmorganca/ollama/app/tray/commontray"
+	"github.com/ollama/ollama/app/tray/commontray"
 )
 
 func InitPlatformTray(icon, updateIcon []byte) (commontray.OllamaTray, error) {

app/tray/tray_windows.go

@@ -1,8 +1,8 @@
 package tray
 
 import (
-	"github.com/jmorganca/ollama/app/tray/commontray"
-	"github.com/jmorganca/ollama/app/tray/wintray"
+	"github.com/ollama/ollama/app/tray/commontray"
+	"github.com/ollama/ollama/app/tray/wintray"
 )
 
 func InitPlatformTray(icon, updateIcon []byte) (commontray.OllamaTray, error) {

app/tray/wintray/eventloop.go

@@ -47,7 +47,6 @@ func nativeLoop() {
 		default:
 			pTranslateMessage.Call(uintptr(unsafe.Pointer(m))) //nolint:errcheck
 			pDispatchMessage.Call(uintptr(unsafe.Pointer(m)))  //nolint:errcheck
-
 		}
 	}
 }
@@ -160,8 +159,8 @@ func (t *winTray) wndProc(hWnd windows.Handle, message uint32, wParam, lParam ui
 		lResult, _, _ = pDefWindowProc.Call(
 			uintptr(hWnd),
 			uintptr(message),
-			uintptr(wParam),
-			uintptr(lParam),
+			wParam,
+			lParam,
 		)
 	}
 	return

app/tray/wintray/menus.go

@@ -1,71 +1,71 @@
-//go:build windows
-
-package wintray
-
-import (
-	"fmt"
-	"log/slog"
-	"unsafe"
-
-	"golang.org/x/sys/windows"
-)
-
-const (
-	updatAvailableMenuID = 1
-	updateMenuID         = updatAvailableMenuID + 1
-	separatorMenuID      = updateMenuID + 1
-	diagLogsMenuID       = separatorMenuID + 1
-	diagSeparatorMenuID  = diagLogsMenuID + 1
-	quitMenuID           = diagSeparatorMenuID + 1
-)
-
-func (t *winTray) initMenus() error {
-	if err := t.addOrUpdateMenuItem(diagLogsMenuID, 0, diagLogsMenuTitle, false); err != nil {
-		return fmt.Errorf("unable to create menu entries %w\n", err)
-	}
-	if err := t.addSeparatorMenuItem(diagSeparatorMenuID, 0); err != nil {
-		return fmt.Errorf("unable to create menu entries %w", err)
-	}
-	if err := t.addOrUpdateMenuItem(quitMenuID, 0, quitMenuTitle, false); err != nil {
-		return fmt.Errorf("unable to create menu entries %w\n", err)
-	}
-	return nil
-}
-
-func (t *winTray) UpdateAvailable(ver string) error {
-	if !t.updateNotified {
-		slog.Debug("updating menu and sending notification for new update")
-		if err := t.addOrUpdateMenuItem(updatAvailableMenuID, 0, updateAvailableMenuTitle, true); err != nil {
-			return fmt.Errorf("unable to create menu entries %w", err)
-		}
-		if err := t.addOrUpdateMenuItem(updateMenuID, 0, updateMenutTitle, false); err != nil {
-			return fmt.Errorf("unable to create menu entries %w", err)
-		}
-		if err := t.addSeparatorMenuItem(separatorMenuID, 0); err != nil {
-			return fmt.Errorf("unable to create menu entries %w", err)
-		}
-		iconFilePath, err := iconBytesToFilePath(wt.updateIcon)
-		if err != nil {
-			return fmt.Errorf("unable to write icon data to temp file: %w", err)
-		}
-		if err := wt.setIcon(iconFilePath); err != nil {
-			return fmt.Errorf("unable to set icon: %w", err)
-		}
-		t.updateNotified = true
-
-		t.pendingUpdate = true
-		// Now pop up the notification
-		t.muNID.Lock()
-		defer t.muNID.Unlock()
-		copy(t.nid.InfoTitle[:], windows.StringToUTF16(updateTitle))
-		copy(t.nid.Info[:], windows.StringToUTF16(fmt.Sprintf(updateMessage, ver)))
-		t.nid.Flags |= NIF_INFO
-		t.nid.Timeout = 10
-		t.nid.Size = uint32(unsafe.Sizeof(*wt.nid))
-		err = t.nid.modify()
-		if err != nil {
-			return err
-		}
-	}
-	return nil
-}
+//go:build windows
+
+package wintray
+
+import (
+	"fmt"
+	"log/slog"
+	"unsafe"
+
+	"golang.org/x/sys/windows"
+)
+
+const (
+	updatAvailableMenuID = 1
+	updateMenuID         = updatAvailableMenuID + 1
+	separatorMenuID      = updateMenuID + 1
+	diagLogsMenuID       = separatorMenuID + 1
+	diagSeparatorMenuID  = diagLogsMenuID + 1
+	quitMenuID           = diagSeparatorMenuID + 1
+)
+
+func (t *winTray) initMenus() error {
+	if err := t.addOrUpdateMenuItem(diagLogsMenuID, 0, diagLogsMenuTitle, false); err != nil {
+		return fmt.Errorf("unable to create menu entries %w\n", err)
+	}
+	if err := t.addSeparatorMenuItem(diagSeparatorMenuID, 0); err != nil {
+		return fmt.Errorf("unable to create menu entries %w", err)
+	}
+	if err := t.addOrUpdateMenuItem(quitMenuID, 0, quitMenuTitle, false); err != nil {
+		return fmt.Errorf("unable to create menu entries %w\n", err)
+	}
+	return nil
+}
+
+func (t *winTray) UpdateAvailable(ver string) error {
+	if !t.updateNotified {
+		slog.Debug("updating menu and sending notification for new update")
+		if err := t.addOrUpdateMenuItem(updatAvailableMenuID, 0, updateAvailableMenuTitle, true); err != nil {
+			return fmt.Errorf("unable to create menu entries %w", err)
+		}
+		if err := t.addOrUpdateMenuItem(updateMenuID, 0, updateMenutTitle, false); err != nil {
+			return fmt.Errorf("unable to create menu entries %w", err)
+		}
+		if err := t.addSeparatorMenuItem(separatorMenuID, 0); err != nil {
+			return fmt.Errorf("unable to create menu entries %w", err)
+		}
+		iconFilePath, err := iconBytesToFilePath(wt.updateIcon)
+		if err != nil {
+			return fmt.Errorf("unable to write icon data to temp file: %w", err)
+		}
+		if err := wt.setIcon(iconFilePath); err != nil {
+			return fmt.Errorf("unable to set icon: %w", err)
+		}
+		t.updateNotified = true
+
+		t.pendingUpdate = true
+		// Now pop up the notification
+		t.muNID.Lock()
+		defer t.muNID.Unlock()
+		copy(t.nid.InfoTitle[:], windows.StringToUTF16(updateTitle))
+		copy(t.nid.Info[:], windows.StringToUTF16(fmt.Sprintf(updateMessage, ver)))
+		t.nid.Flags |= NIF_INFO
+		t.nid.Timeout = 10
+		t.nid.Size = uint32(unsafe.Sizeof(*wt.nid))
+		err = t.nid.modify()
+		if err != nil {
+			return err
+		}
+	}
+	return nil
+}

app/tray/wintray/tray.go

@@ -13,7 +13,7 @@ import (
 	"sync"
 	"unsafe"
 
-	"github.com/jmorganca/ollama/app/tray/commontray"
+	"github.com/ollama/ollama/app/tray/commontray"
 	"golang.org/x/sys/windows"
 )
 
@@ -186,7 +186,7 @@ func (t *winTray) initInstance() error {
 	t.muNID.Lock()
 	defer t.muNID.Unlock()
 	t.nid = &notifyIconData{
-		Wnd:             windows.Handle(t.window),
+		Wnd:             t.window,
 		ID:              100,
 		Flags:           NIF_MESSAGE,
 		CallbackMessage: t.wmSystrayMessage,
@@ -197,7 +197,6 @@ func (t *winTray) initInstance() error {
 }
 
 func (t *winTray) createMenu() error {
-
 	menuHandle, _, err := pCreatePopupMenu.Call()
 	if menuHandle == 0 {
 		return err
@@ -246,7 +245,7 @@ func (t *winTray) addOrUpdateMenuItem(menuItemId uint32, parentId uint32, title
 	mi := menuItemInfo{
 		Mask:     MIIM_FTYPE | MIIM_STRING | MIIM_ID | MIIM_STATE,
 		Type:     MFT_STRING,
-		ID:       uint32(menuItemId),
+		ID:       menuItemId,
 		TypeData: titlePtr,
 		Cch:      uint32(len(title)),
 	}
@@ -302,11 +301,10 @@ func (t *winTray) addOrUpdateMenuItem(menuItemId uint32, parentId uint32, title
 }
 
 func (t *winTray) addSeparatorMenuItem(menuItemId, parentId uint32) error {
-
 	mi := menuItemInfo{
 		Mask: MIIM_FTYPE | MIIM_ID | MIIM_STATE,
 		Type: MFT_SEPARATOR,
-		ID:   uint32(menuItemId),
+		ID:   menuItemId,
 	}
 
 	mi.Size = uint32(unsafe.Sizeof(mi))
@@ -426,7 +424,6 @@ func iconBytesToFilePath(iconBytes []byte) (string, error) {
 // Loads an image from file and shows it in tray.
 // Shell_NotifyIcon: https://msdn.microsoft.com/en-us/library/windows/desktop/bb762159(v=vs.85).aspx
 func (t *winTray) setIcon(src string) error {
-
 	h, err := t.loadIconFrom(src)
 	if err != nil {
 		return err
@@ -444,7 +441,6 @@ func (t *winTray) setIcon(src string) error {
 // Loads an image from file to be shown in tray or menu item.
 // LoadImage: https://msdn.microsoft.com/en-us/library/windows/desktop/ms648045(v=vs.85).aspx
 func (t *winTray) loadIconFrom(src string) (windows.Handle, error) {
-
 	// Save and reuse handles of loaded images
 	t.muLoadedImages.RLock()
 	h, ok := t.loadedImages[src]

auth/auth.go

@@ -10,12 +10,44 @@ import (
 	"log/slog"
 	"os"
 	"path/filepath"
+	"strings"
 
 	"golang.org/x/crypto/ssh"
 )
 
 const defaultPrivateKey = "id_ed25519"
 
+func keyPath() (string, error) {
+	home, err := os.UserHomeDir()
+	if err != nil {
+		return "", err
+	}
+
+	return filepath.Join(home, ".ollama", defaultPrivateKey), nil
+}
+
+func GetPublicKey() (string, error) {
+	keyPath, err := keyPath()
+	if err != nil {
+		return "", err
+	}
+
+	privateKeyFile, err := os.ReadFile(keyPath)
+	if err != nil {
+		slog.Info(fmt.Sprintf("Failed to load private key: %v", err))
+		return "", err
+	}
+
+	privateKey, err := ssh.ParsePrivateKey(privateKeyFile)
+	if err != nil {
+		return "", err
+	}
+
+	publicKey := ssh.MarshalAuthorizedKey(privateKey.PublicKey())
+
+	return strings.TrimSpace(string(publicKey)), nil
+}
+
 func NewNonce(r io.Reader, length int) (string, error) {
 	nonce := make([]byte, length)
 	if _, err := io.ReadFull(r, nonce); err != nil {
@@ -26,13 +58,11 @@ func NewNonce(r io.Reader, length int) (string, error) {
 }
 
 func Sign(ctx context.Context, bts []byte) (string, error) {
-	home, err := os.UserHomeDir()
+	keyPath, err := keyPath()
 	if err != nil {
 		return "", err
 	}
 
-	keyPath := filepath.Join(home, ".ollama", defaultPrivateKey)
-
 	privateKeyFile, err := os.ReadFile(keyPath)
 	if err != nil {
 		slog.Info(fmt.Sprintf("Failed to load private key: %v", err))

cmd/cmd.go

@@ -1,6 +1,7 @@
 package cmd
 
 import (
+	"archive/zip"
 	"bytes"
 	"context"
 	"crypto/ed25519"
@@ -11,30 +12,36 @@ import (
 	"fmt"
 	"io"
 	"log"
+	"math"
 	"net"
 	"net/http"
 	"os"
 	"os/signal"
 	"path/filepath"
+	"regexp"
 	"runtime"
+	"slices"
 	"strings"
 	"syscall"
 	"time"
 
 	"github.com/containerd/console"
-
+	"github.com/mattn/go-runewidth"
 	"github.com/olekukonko/tablewriter"
 	"github.com/spf13/cobra"
 	"golang.org/x/crypto/ssh"
-	"golang.org/x/exp/slices"
 	"golang.org/x/term"
 
-	"github.com/jmorganca/ollama/api"
-	"github.com/jmorganca/ollama/format"
-	"github.com/jmorganca/ollama/parser"
-	"github.com/jmorganca/ollama/progress"
-	"github.com/jmorganca/ollama/server"
-	"github.com/jmorganca/ollama/version"
+	"github.com/ollama/ollama/api"
+	"github.com/ollama/ollama/auth"
+	"github.com/ollama/ollama/envconfig"
+	"github.com/ollama/ollama/format"
+	"github.com/ollama/ollama/parser"
+	"github.com/ollama/ollama/progress"
+	"github.com/ollama/ollama/server"
+	"github.com/ollama/ollama/types/errtypes"
+	"github.com/ollama/ollama/types/model"
+	"github.com/ollama/ollama/version"
 )
 
 func CreateHandler(cmd *cobra.Command, args []string) error {
@@ -52,14 +59,13 @@ func CreateHandler(cmd *cobra.Command, args []string) error {
 	p := progress.NewProgress(os.Stderr)
 	defer p.Stop()
 
-	bars := make(map[string]*progress.Bar)
-
-	modelfile, err := os.ReadFile(filename)
+	f, err := os.Open(filename)
 	if err != nil {
 		return err
 	}
+	defer f.Close()
 
-	commands, err := parser.Parse(bytes.NewReader(modelfile))
+	modelfile, err := parser.ParseFile(f)
 	if err != nil {
 		return err
 	}
@@ -73,10 +79,10 @@ func CreateHandler(cmd *cobra.Command, args []string) error {
 	spinner := progress.NewSpinner(status)
 	p.Add(status, spinner)
 
-	for _, c := range commands {
-		switch c.Name {
+	for i := range modelfile.Commands {
+		switch modelfile.Commands[i].Name {
 		case "model", "adapter":
-			path := c.Args
+			path := modelfile.Commands[i].Args
 			if path == "~" {
 				path = home
 			} else if strings.HasPrefix(path, "~/") {
@@ -87,29 +93,35 @@ func CreateHandler(cmd *cobra.Command, args []string) error {
 				path = filepath.Join(filepath.Dir(filename), path)
 			}
 
-			bin, err := os.Open(path)
-			if errors.Is(err, os.ErrNotExist) && c.Name == "model" {
+			fi, err := os.Stat(path)
+			if errors.Is(err, os.ErrNotExist) && modelfile.Commands[i].Name == "model" {
 				continue
 			} else if err != nil {
 				return err
 			}
-			defer bin.Close()
 
-			hash := sha256.New()
-			if _, err := io.Copy(hash, bin); err != nil {
-				return err
+			if fi.IsDir() {
+				// this is likely a safetensors or pytorch directory
+				// TODO make this work w/ adapters
+				tempfile, err := tempZipFiles(path)
+				if err != nil {
+					return err
+				}
+				defer os.RemoveAll(tempfile)
+
+				path = tempfile
 			}
-			bin.Seek(0, io.SeekStart)
 
-			digest := fmt.Sprintf("sha256:%x", hash.Sum(nil))
-			if err = client.CreateBlob(cmd.Context(), digest, bin); err != nil {
+			digest, err := createBlob(cmd, client, path)
+			if err != nil {
 				return err
 			}
 
-			modelfile = bytes.ReplaceAll(modelfile, []byte(c.Args), []byte("@"+digest))
+			modelfile.Commands[i].Args = "@" + digest
 		}
 	}
 
+	bars := make(map[string]*progress.Bar)
 	fn := func(resp api.ProgressResponse) error {
 		if resp.Digest != "" {
 			spinner.Stop()
@@ -133,7 +145,9 @@ func CreateHandler(cmd *cobra.Command, args []string) error {
 		return nil
 	}
 
-	request := api.CreateRequest{Name: args[0], Modelfile: string(modelfile)}
+	quantize, _ := cmd.Flags().GetString("quantize")
+
+	request := api.CreateRequest{Name: args[0], Modelfile: modelfile.String(), Quantize: quantize}
 	if err := client.Create(cmd.Context(), &request, fn); err != nil {
 		return err
 	}
@@ -141,39 +155,144 @@ func CreateHandler(cmd *cobra.Command, args []string) error {
 	return nil
 }
 
-func RunHandler(cmd *cobra.Command, args []string) error {
-	client, err := api.ClientFromEnvironment()
+func tempZipFiles(path string) (string, error) {
+	tempfile, err := os.CreateTemp("", "ollama-tf")
 	if err != nil {
-		return err
+		return "", err
 	}
+	defer tempfile.Close()
 
-	name := args[0]
-
-	// check if the model exists on the server
-	show, err := client.Show(cmd.Context(), &api.ShowRequest{Name: name})
-	var statusError api.StatusError
-	switch {
-	case errors.As(err, &statusError) && statusError.StatusCode == http.StatusNotFound:
-		if err := PullHandler(cmd, []string{name}); err != nil {
-			return err
-		}
-
-		show, err = client.Show(cmd.Context(), &api.ShowRequest{Name: name})
+	detectContentType := func(path string) (string, error) {
+		f, err := os.Open(path)
 		if err != nil {
-			return err
+			return "", err
 		}
-	case err != nil:
-		return err
+		defer f.Close()
+
+		var b bytes.Buffer
+		b.Grow(512)
+
+		if _, err := io.CopyN(&b, f, 512); err != nil && !errors.Is(err, io.EOF) {
+			return "", err
+		}
+
+		contentType, _, _ := strings.Cut(http.DetectContentType(b.Bytes()), ";")
+		return contentType, nil
 	}
 
+	glob := func(pattern, contentType string) ([]string, error) {
+		matches, err := filepath.Glob(pattern)
+		if err != nil {
+			return nil, err
+		}
+
+		for _, safetensor := range matches {
+			if ct, err := detectContentType(safetensor); err != nil {
+				return nil, err
+			} else if ct != contentType {
+				return nil, fmt.Errorf("invalid content type: expected %s for %s", ct, safetensor)
+			}
+		}
+
+		return matches, nil
+	}
+
+	var files []string
+	if st, _ := glob(filepath.Join(path, "model*.safetensors"), "application/octet-stream"); len(st) > 0 {
+		// safetensors files might be unresolved git lfs references; skip if they are
+		// covers model-x-of-y.safetensors, model.fp32-x-of-y.safetensors, model.safetensors
+		files = append(files, st...)
+	} else if pt, _ := glob(filepath.Join(path, "pytorch_model*.bin"), "application/zip"); len(pt) > 0 {
+		// pytorch files might also be unresolved git lfs references; skip if they are
+		// covers pytorch_model-x-of-y.bin, pytorch_model.fp32-x-of-y.bin, pytorch_model.bin
+		files = append(files, pt...)
+	} else if pt, _ := glob(filepath.Join(path, "consolidated*.pth"), "application/zip"); len(pt) > 0 {
+		// pytorch files might also be unresolved git lfs references; skip if they are
+		// covers consolidated.x.pth, consolidated.pth
+		files = append(files, pt...)
+	} else {
+		return "", errors.New("no safetensors or torch files found")
+	}
+
+	// add configuration files, json files are detected as text/plain
+	js, err := glob(filepath.Join(path, "*.json"), "text/plain")
+	if err != nil {
+		return "", err
+	}
+	files = append(files, js...)
+
+	if tks, _ := glob(filepath.Join(path, "tokenizer.model"), "application/octet-stream"); len(tks) > 0 {
+		// add tokenizer.model if it exists, tokenizer.json is automatically picked up by the previous glob
+		// tokenizer.model might be a unresolved git lfs reference; error if it is
+		files = append(files, tks...)
+	} else if tks, _ := glob(filepath.Join(path, "**/tokenizer.model"), "text/plain"); len(tks) > 0 {
+		// some times tokenizer.model is in a subdirectory (e.g. meta-llama/Meta-Llama-3-8B)
+		files = append(files, tks...)
+	}
+
+	zipfile := zip.NewWriter(tempfile)
+	defer zipfile.Close()
+
+	for _, file := range files {
+		f, err := os.Open(file)
+		if err != nil {
+			return "", err
+		}
+		defer f.Close()
+
+		fi, err := f.Stat()
+		if err != nil {
+			return "", err
+		}
+
+		zfi, err := zip.FileInfoHeader(fi)
+		if err != nil {
+			return "", err
+		}
+
+		zf, err := zipfile.CreateHeader(zfi)
+		if err != nil {
+			return "", err
+		}
+
+		if _, err := io.Copy(zf, f); err != nil {
+			return "", err
+		}
+	}
+
+	return tempfile.Name(), nil
+}
+
+func createBlob(cmd *cobra.Command, client *api.Client, path string) (string, error) {
+	bin, err := os.Open(path)
+	if err != nil {
+		return "", err
+	}
+	defer bin.Close()
+
+	hash := sha256.New()
+	if _, err := io.Copy(hash, bin); err != nil {
+		return "", err
+	}
+
+	if _, err := bin.Seek(0, io.SeekStart); err != nil {
+		return "", err
+	}
+
+	digest := fmt.Sprintf("sha256:%x", hash.Sum(nil))
+	if err = client.CreateBlob(cmd.Context(), digest, bin); err != nil {
+		return "", err
+	}
+	return digest, nil
+}
+
+func RunHandler(cmd *cobra.Command, args []string) error {
 	interactive := true
 
 	opts := runOptions{
-		Model:       args[0],
-		WordWrap:    os.Getenv("TERM") == "xterm-256color",
-		Options:     map[string]interface{}{},
-		MultiModal:  slices.Contains(show.Details.Families, "clip"),
-		ParentModel: show.Details.ParentModel,
+		Model:    args[0],
+		WordWrap: os.Getenv("TERM") == "xterm-256color",
+		Options:  map[string]interface{}{},
 	}
 
 	format, err := cmd.Flags().GetString("format")
@@ -182,6 +301,18 @@ func RunHandler(cmd *cobra.Command, args []string) error {
 	}
 	opts.Format = format
 
+	keepAlive, err := cmd.Flags().GetString("keepalive")
+	if err != nil {
+		return err
+	}
+	if keepAlive != "" {
+		d, err := time.ParseDuration(keepAlive)
+		if err != nil {
+			return err
+		}
+		opts.KeepAlive = &api.Duration{Duration: d}
+	}
+
 	prompts := args[1:]
 	// prepend stdin to the prompt if provided
 	if !term.IsTerminal(int(os.Stdin.Fd())) {
@@ -205,11 +336,79 @@ func RunHandler(cmd *cobra.Command, args []string) error {
 	}
 	opts.WordWrap = !nowrap
 
-	if !interactive {
-		return generate(cmd, opts)
+	// Fill out the rest of the options based on information about the
+	// model.
+	client, err := api.ClientFromEnvironment()
+	if err != nil {
+		return err
 	}
 
-	return generateInteractive(cmd, opts)
+	name := args[0]
+	info, err := func() (*api.ShowResponse, error) {
+		showReq := &api.ShowRequest{Name: name}
+		info, err := client.Show(cmd.Context(), showReq)
+		var se api.StatusError
+		if errors.As(err, &se) && se.StatusCode == http.StatusNotFound {
+			if err := PullHandler(cmd, []string{name}); err != nil {
+				return nil, err
+			}
+			return client.Show(cmd.Context(), &api.ShowRequest{Name: name})
+		}
+		return info, err
+	}()
+	if err != nil {
+		return err
+	}
+
+	opts.MultiModal = slices.Contains(info.Details.Families, "clip")
+	opts.ParentModel = info.Details.ParentModel
+	opts.Messages = append(opts.Messages, info.Messages...)
+
+	if interactive {
+		return generateInteractive(cmd, opts)
+	}
+	return generate(cmd, opts)
+}
+
+func errFromUnknownKey(unknownKeyErr error) error {
+	// find SSH public key in the error message
+	sshKeyPattern := `ssh-\w+ [^\s"]+`
+	re := regexp.MustCompile(sshKeyPattern)
+	matches := re.FindStringSubmatch(unknownKeyErr.Error())
+
+	if len(matches) > 0 {
+		serverPubKey := matches[0]
+
+		localPubKey, err := auth.GetPublicKey()
+		if err != nil {
+			return unknownKeyErr
+		}
+
+		if runtime.GOOS == "linux" && serverPubKey != localPubKey {
+			// try the ollama service public key
+			svcPubKey, err := os.ReadFile("/usr/share/ollama/.ollama/id_ed25519.pub")
+			if err != nil {
+				return unknownKeyErr
+			}
+			localPubKey = strings.TrimSpace(string(svcPubKey))
+		}
+
+		// check if the returned public key matches the local public key, this prevents adding a remote key to the user's account
+		if serverPubKey != localPubKey {
+			return unknownKeyErr
+		}
+
+		var msg strings.Builder
+		msg.WriteString(unknownKeyErr.Error())
+		msg.WriteString("\n\nYour ollama key is:\n")
+		msg.WriteString(localPubKey)
+		msg.WriteString("\nAdd your key at:\n")
+		msg.WriteString("https://ollama.com/settings/keys")
+
+		return errors.New(msg.String())
+	}
+
+	return unknownKeyErr
 }
 
 func PushHandler(cmd *cobra.Command, args []string) error {
@@ -259,6 +458,20 @@ func PushHandler(cmd *cobra.Command, args []string) error {
 
 	request := api.PushRequest{Name: args[0], Insecure: insecure}
 	if err := client.Push(cmd.Context(), &request, fn); err != nil {
+		if spinner != nil {
+			spinner.Stop()
+		}
+		if strings.Contains(err.Error(), "access denied") {
+			return errors.New("you are not authorized to push to this namespace, create the model under a namespace you own")
+		}
+		host := model.ParseName(args[0]).Host
+		isOllamaHost := strings.HasSuffix(host, ".ollama.ai") || strings.HasSuffix(host, ".ollama.com")
+		if strings.Contains(err.Error(), errtypes.UnknownOllamaKeyErrMsg) && isOllamaHost {
+			// the user has not added their ollama key to ollama.com
+			// re-throw an error with a more user-friendly message
+			return errFromUnknownKey(err)
+		}
+
 		return err
 	}
 
@@ -299,6 +512,52 @@ func ListHandler(cmd *cobra.Command, args []string) error {
 	return nil
 }
 
+func ListRunningHandler(cmd *cobra.Command, args []string) error {
+	client, err := api.ClientFromEnvironment()
+	if err != nil {
+		return err
+	}
+
+	models, err := client.ListRunning(cmd.Context())
+	if err != nil {
+		return err
+	}
+
+	var data [][]string
+
+	for _, m := range models.Models {
+		if len(args) == 0 || strings.HasPrefix(m.Name, args[0]) {
+			var procStr string
+			switch {
+			case m.SizeVRAM == 0:
+				procStr = "100% CPU"
+			case m.SizeVRAM == m.Size:
+				procStr = "100% GPU"
+			case m.SizeVRAM > m.Size || m.Size == 0:
+				procStr = "Unknown"
+			default:
+				sizeCPU := m.Size - m.SizeVRAM
+				cpuPercent := math.Round(float64(sizeCPU) / float64(m.Size) * 100)
+				procStr = fmt.Sprintf("%d%%/%d%% CPU/GPU", int(cpuPercent), int(100-cpuPercent))
+			}
+			data = append(data, []string{m.Name, m.Digest[:12], format.HumanBytes(m.Size), procStr, format.HumanTime(m.ExpiresAt, "Never")})
+		}
+	}
+
+	table := tablewriter.NewWriter(os.Stdout)
+	table.SetHeader([]string{"NAME", "ID", "SIZE", "PROCESSOR", "UNTIL"})
+	table.SetHeaderAlignment(tablewriter.ALIGN_LEFT)
+	table.SetAlignment(tablewriter.ALIGN_LEFT)
+	table.SetHeaderLine(false)
+	table.SetBorder(false)
+	table.SetNoWhiteSpace(true)
+	table.SetTablePadding("\t")
+	table.AppendBulk(data)
+	table.Render()
+
+	return nil
+}
+
 func DeleteHandler(cmd *cobra.Command, args []string) error {
 	client, err := api.ClientFromEnvironment()
 	if err != nil {
@@ -321,10 +580,6 @@ func ShowHandler(cmd *cobra.Command, args []string) error {
 		return err
 	}
 
-	if len(args) != 1 {
-		return errors.New("missing model name")
-	}
-
 	license, errLicense := cmd.Flags().GetBool("license")
 	modelfile, errModelfile := cmd.Flags().GetBool("modelfile")
 	parameters, errParams := cmd.Flags().GetBool("parameters")
@@ -367,8 +622,6 @@ func ShowHandler(cmd *cobra.Command, args []string) error {
 
 	if flagsSet > 1 {
 		return errors.New("only one of '--license', '--modelfile', '--parameters', '--system', or '--template' can be specified")
-	} else if flagsSet == 0 {
-		return errors.New("one of '--license', '--modelfile', '--parameters', '--system', or '--template' must be specified")
 	}
 
 	req := api.ShowRequest{Name: args[0]}
@@ -377,22 +630,141 @@ func ShowHandler(cmd *cobra.Command, args []string) error {
 		return err
 	}
 
-	switch showType {
-	case "license":
-		fmt.Println(resp.License)
-	case "modelfile":
-		fmt.Println(resp.Modelfile)
-	case "parameters":
-		fmt.Println(resp.Parameters)
-	case "system":
-		fmt.Println(resp.System)
-	case "template":
-		fmt.Println(resp.Template)
+	if flagsSet == 1 {
+		switch showType {
+		case "license":
+			fmt.Println(resp.License)
+		case "modelfile":
+			fmt.Println(resp.Modelfile)
+		case "parameters":
+			fmt.Println(resp.Parameters)
+		case "system":
+			fmt.Println(resp.System)
+		case "template":
+			fmt.Println(resp.Template)
+		}
+
+		return nil
 	}
 
+	showInfo(resp)
+
 	return nil
 }
 
+func showInfo(resp *api.ShowResponse) {
+	arch := resp.ModelInfo["general.architecture"].(string)
+
+	modelData := [][]string{
+		{"arch", arch},
+		{"parameters", resp.Details.ParameterSize},
+		{"quantization", resp.Details.QuantizationLevel},
+		{"context length", fmt.Sprintf("%v", resp.ModelInfo[fmt.Sprintf("%s.context_length", arch)].(float64))},
+		{"embedding length", fmt.Sprintf("%v", resp.ModelInfo[fmt.Sprintf("%s.embedding_length", arch)].(float64))},
+	}
+
+	mainTableData := [][]string{
+		{"Model"},
+		{renderSubTable(modelData, false)},
+	}
+
+	if resp.ProjectorInfo != nil {
+		projectorData := [][]string{
+			{"arch", "clip"},
+			{"parameters", format.HumanNumber(uint64(resp.ProjectorInfo["general.parameter_count"].(float64)))},
+		}
+
+		if projectorType, ok := resp.ProjectorInfo["clip.projector_type"]; ok {
+			projectorData = append(projectorData, []string{"projector type", projectorType.(string)})
+		}
+
+		projectorData = append(projectorData,
+			[]string{"embedding length", fmt.Sprintf("%v", resp.ProjectorInfo["clip.vision.embedding_length"].(float64))},
+			[]string{"projection dimensionality", fmt.Sprintf("%v", resp.ProjectorInfo["clip.vision.projection_dim"].(float64))},
+		)
+
+		mainTableData = append(mainTableData,
+			[]string{"Projector"},
+			[]string{renderSubTable(projectorData, false)},
+		)
+	}
+
+	if resp.Parameters != "" {
+		mainTableData = append(mainTableData, []string{"Parameters"}, []string{formatParams(resp.Parameters)})
+	}
+
+	if resp.System != "" {
+		mainTableData = append(mainTableData, []string{"System"}, []string{renderSubTable(twoLines(resp.System), true)})
+	}
+
+	if resp.License != "" {
+		mainTableData = append(mainTableData, []string{"License"}, []string{renderSubTable(twoLines(resp.License), true)})
+	}
+
+	table := tablewriter.NewWriter(os.Stdout)
+	table.SetAutoWrapText(false)
+	table.SetBorder(false)
+	table.SetAlignment(tablewriter.ALIGN_LEFT)
+
+	for _, v := range mainTableData {
+		table.Append(v)
+	}
+
+	table.Render()
+}
+
+func renderSubTable(data [][]string, file bool) string {
+	var buf bytes.Buffer
+	table := tablewriter.NewWriter(&buf)
+	table.SetAutoWrapText(!file)
+	table.SetBorder(false)
+	table.SetNoWhiteSpace(true)
+	table.SetTablePadding("\t")
+	table.SetAlignment(tablewriter.ALIGN_LEFT)
+
+	for _, v := range data {
+		table.Append(v)
+	}
+
+	table.Render()
+
+	renderedTable := buf.String()
+	lines := strings.Split(renderedTable, "\n")
+	for i, line := range lines {
+		lines[i] = "\t" + line
+	}
+
+	return strings.Join(lines, "\n")
+}
+
+func twoLines(s string) [][]string {
+	lines := strings.Split(s, "\n")
+	res := [][]string{}
+
+	count := 0
+	for _, line := range lines {
+		line = strings.TrimSpace(line)
+		if line != "" {
+			count++
+			res = append(res, []string{line})
+			if count == 2 {
+				return res
+			}
+		}
+	}
+	return res
+}
+
+func formatParams(s string) string {
+	lines := strings.Split(s, "\n")
+	table := [][]string{}
+
+	for _, line := range lines {
+		table = append(table, strings.Fields(line))
+	}
+	return renderSubTable(table, false)
+}
+
 func CopyHandler(cmd *cobra.Command, args []string) error {
 	client, err := api.ClientFromEnvironment()
 	if err != nil {
@@ -471,10 +843,10 @@ type runOptions struct {
 	WordWrap    bool
 	Format      string
 	System      string
-	Template    string
 	Images      []api.ImageData
 	Options     map[string]interface{}
 	MultiModal  bool
+	KeepAlive   *api.Duration
 }
 
 type displayResponseState struct {
@@ -487,7 +859,7 @@ func displayResponse(content string, wordWrap bool, state *displayResponseState)
 	if wordWrap && termWidth >= 10 {
 		for _, ch := range content {
 			if state.lineLength+1 > termWidth-5 {
-				if len(state.wordBuffer) > termWidth-10 {
+				if runewidth.StringWidth(state.wordBuffer) > termWidth-10 {
 					fmt.Printf("%s%c", state.wordBuffer, ch)
 					state.wordBuffer = ""
 					state.lineLength = 0
@@ -495,12 +867,22 @@ func displayResponse(content string, wordWrap bool, state *displayResponseState)
 				}
 
 				// backtrack the length of the last word and clear to the end of the line
-				fmt.Printf("\x1b[%dD\x1b[K\n", len(state.wordBuffer))
+				a := runewidth.StringWidth(state.wordBuffer)
+				if a > 0 {
+					fmt.Printf("\x1b[%dD", a)
+				}
+				fmt.Printf("\x1b[K\n")
 				fmt.Printf("%s%c", state.wordBuffer, ch)
-				state.lineLength = len(state.wordBuffer) + 1
+				chWidth := runewidth.RuneWidth(ch)
+
+				state.lineLength = runewidth.StringWidth(state.wordBuffer) + chWidth
 			} else {
 				fmt.Print(string(ch))
-				state.lineLength += 1
+				state.lineLength += runewidth.RuneWidth(ch)
+				if runewidth.RuneWidth(ch) >= 2 {
+					state.wordBuffer = ""
+					continue
+				}
 
 				switch ch {
 				case ' ':
@@ -569,6 +951,10 @@ func chat(cmd *cobra.Command, opts runOptions) (*api.Message, error) {
 		Options:  opts.Options,
 	}
 
+	if opts.KeepAlive != nil {
+		req.KeepAlive = opts.KeepAlive
+	}
+
 	if err := client.Chat(cancelCtx, req, fn); err != nil {
 		if errors.Is(err, context.Canceled) {
 			return nil, nil
@@ -644,14 +1030,14 @@ func generate(cmd *cobra.Command, opts runOptions) error {
 	}
 
 	request := api.GenerateRequest{
-		Model:    opts.Model,
-		Prompt:   opts.Prompt,
-		Context:  generateContext,
-		Images:   opts.Images,
-		Format:   opts.Format,
-		System:   opts.System,
-		Template: opts.Template,
-		Options:  opts.Options,
+		Model:     opts.Model,
+		Prompt:    opts.Prompt,
+		Context:   generateContext,
+		Images:    opts.Images,
+		Format:    opts.Format,
+		System:    opts.System,
+		Options:   opts.Options,
+		KeepAlive: opts.KeepAlive,
 	}
 
 	if err := client.Generate(ctx, &request, fn); err != nil {
@@ -686,24 +1072,21 @@ func generate(cmd *cobra.Command, opts runOptions) error {
 }
 
 func RunServer(cmd *cobra.Command, _ []string) error {
-	host, port, err := net.SplitHostPort(strings.Trim(os.Getenv("OLLAMA_HOST"), "\"'"))
-	if err != nil {
-		host, port = "127.0.0.1", "11434"
-		if ip := net.ParseIP(strings.Trim(os.Getenv("OLLAMA_HOST"), "[]")); ip != nil {
-			host = ip.String()
-		}
-	}
-
 	if err := initializeKeypair(); err != nil {
 		return err
 	}
 
-	ln, err := net.Listen("tcp", net.JoinHostPort(host, port))
+	ln, err := net.Listen("tcp", net.JoinHostPort(envconfig.Host.Host, envconfig.Host.Port))
 	if err != nil {
 		return err
 	}
 
-	return server.Serve(ln)
+	err = server.Serve(ln)
+	if errors.Is(err, http.ErrServerClosed) {
+		return nil
+	}
+
+	return err
 }
 
 func initializeKeypair() error {
@@ -718,61 +1101,40 @@ func initializeKeypair() error {
 	_, err = os.Stat(privKeyPath)
 	if os.IsNotExist(err) {
 		fmt.Printf("Couldn't find '%s'. Generating new private key.\n", privKeyPath)
-		_, privKey, err := ed25519.GenerateKey(rand.Reader)
+		cryptoPublicKey, cryptoPrivateKey, err := ed25519.GenerateKey(rand.Reader)
 		if err != nil {
 			return err
 		}
 
-		privKeyBytes, err := format.OpenSSHPrivateKey(privKey, "")
+		privateKeyBytes, err := ssh.MarshalPrivateKey(cryptoPrivateKey, "")
 		if err != nil {
 			return err
 		}
 
-		err = os.MkdirAll(filepath.Dir(privKeyPath), 0o755)
-		if err != nil {
+		if err := os.MkdirAll(filepath.Dir(privKeyPath), 0o755); err != nil {
 			return fmt.Errorf("could not create directory %w", err)
 		}
 
-		err = os.WriteFile(privKeyPath, pem.EncodeToMemory(privKeyBytes), 0o600)
+		if err := os.WriteFile(privKeyPath, pem.EncodeToMemory(privateKeyBytes), 0o600); err != nil {
+			return err
+		}
+
+		sshPublicKey, err := ssh.NewPublicKey(cryptoPublicKey)
 		if err != nil {
 			return err
 		}
 
-		sshPrivateKey, err := ssh.NewSignerFromKey(privKey)
-		if err != nil {
+		publicKeyBytes := ssh.MarshalAuthorizedKey(sshPublicKey)
+
+		if err := os.WriteFile(pubKeyPath, publicKeyBytes, 0o644); err != nil {
 			return err
 		}
 
-		pubKeyData := ssh.MarshalAuthorizedKey(sshPrivateKey.PublicKey())
-
-		err = os.WriteFile(pubKeyPath, pubKeyData, 0o644)
-		if err != nil {
-			return err
-		}
-
-		fmt.Printf("Your new public key is: \n\n%s\n", string(pubKeyData))
+		fmt.Printf("Your new public key is: \n\n%s\n", publicKeyBytes)
 	}
 	return nil
 }
 
-//nolint:unused
-func waitForServer(ctx context.Context, client *api.Client) error {
-	// wait for the server to start
-	timeout := time.After(5 * time.Second)
-	tick := time.Tick(500 * time.Millisecond)
-	for {
-		select {
-		case <-timeout:
-			return errors.New("timed out waiting for server to start")
-		case <-tick:
-			if err := client.Heartbeat(ctx); err == nil {
-				return nil // server has started
-			}
-		}
-	}
-
-}
-
 func checkServerHeartbeat(cmd *cobra.Command, _ []string) error {
 	client, err := api.ClientFromEnvironment()
 	if err != nil {
@@ -809,13 +1171,27 @@ func versionHandler(cmd *cobra.Command, _ []string) {
 	}
 }
 
+func appendEnvDocs(cmd *cobra.Command, envs []envconfig.EnvVar) {
+	if len(envs) == 0 {
+		return
+	}
+
+	envUsage := `
+Environment Variables:
+`
+	for _, e := range envs {
+		envUsage += fmt.Sprintf("      %-24s   %s\n", e.Name, e.Description)
+	}
+
+	cmd.SetUsageTemplate(cmd.UsageTemplate() + envUsage)
+}
+
 func NewCLI() *cobra.Command {
 	log.SetFlags(log.LstdFlags | log.Lshortfile)
 	cobra.EnableCommandSorting = false
 
 	if runtime.GOOS == "windows" {
-		// Enable colorful ANSI escape code in Windows terminal (disabled by default)
-		console.ConsoleFromFile(os.Stdout) //nolint:errcheck
+		console.ConsoleFromFile(os.Stdin) //nolint:errcheck
 	}
 
 	rootCmd := &cobra.Command{
@@ -846,7 +1222,8 @@ func NewCLI() *cobra.Command {
 		RunE:    CreateHandler,
 	}
 
-	createCmd.Flags().StringP("file", "f", "Modelfile", "Name of the Modelfile (default \"Modelfile\")")
+	createCmd.Flags().StringP("file", "f", "Modelfile", "Name of the Modelfile")
+	createCmd.Flags().StringP("quantize", "q", "", "Quantize model to this level (e.g. q4_0)")
 
 	showCmd := &cobra.Command{
 		Use:     "show MODEL",
@@ -870,11 +1247,11 @@ func NewCLI() *cobra.Command {
 		RunE:    RunHandler,
 	}
 
+	runCmd.Flags().String("keepalive", "", "Duration to keep a model loaded (e.g. 5m)")
 	runCmd.Flags().Bool("verbose", false, "Show timings for response")
 	runCmd.Flags().Bool("insecure", false, "Use an insecure registry")
 	runCmd.Flags().Bool("nowordwrap", false, "Don't wrap words to the next line automatically")
 	runCmd.Flags().String("format", "", "Response format (e.g. json)")
-
 	serveCmd := &cobra.Command{
 		Use:     "serve",
 		Aliases: []string{"start"},
@@ -911,8 +1288,15 @@ func NewCLI() *cobra.Command {
 		RunE:    ListHandler,
 	}
 
+	psCmd := &cobra.Command{
+		Use:     "ps",
+		Short:   "List running models",
+		PreRunE: checkServerHeartbeat,
+		RunE:    ListRunningHandler,
+	}
+
 	copyCmd := &cobra.Command{
-		Use:     "cp SOURCE TARGET",
+		Use:     "cp SOURCE DESTINATION",
 		Short:   "Copy a model",
 		Args:    cobra.ExactArgs(2),
 		PreRunE: checkServerHeartbeat,
@@ -927,6 +1311,46 @@ func NewCLI() *cobra.Command {
 		RunE:    DeleteHandler,
 	}
 
+	envVars := envconfig.AsMap()
+
+	envs := []envconfig.EnvVar{envVars["OLLAMA_HOST"]}
+
+	for _, cmd := range []*cobra.Command{
+		createCmd,
+		showCmd,
+		runCmd,
+		pullCmd,
+		pushCmd,
+		listCmd,
+		psCmd,
+		copyCmd,
+		deleteCmd,
+		serveCmd,
+	} {
+		switch cmd {
+		case runCmd:
+			appendEnvDocs(cmd, []envconfig.EnvVar{envVars["OLLAMA_HOST"], envVars["OLLAMA_NOHISTORY"]})
+		case serveCmd:
+			appendEnvDocs(cmd, []envconfig.EnvVar{
+				envVars["OLLAMA_DEBUG"],
+				envVars["OLLAMA_HOST"],
+				envVars["OLLAMA_KEEP_ALIVE"],
+				envVars["OLLAMA_MAX_LOADED_MODELS"],
+				envVars["OLLAMA_MAX_QUEUE"],
+				envVars["OLLAMA_MODELS"],
+				envVars["OLLAMA_NUM_PARALLEL"],
+				envVars["OLLAMA_NOPRUNE"],
+				envVars["OLLAMA_ORIGINS"],
+				envVars["OLLAMA_TMPDIR"],
+				envVars["OLLAMA_FLASH_ATTENTION"],
+				envVars["OLLAMA_LLM_LIBRARY"],
+				envVars["OLLAMA_MAX_VRAM"],
+			})
+		default:
+			appendEnvDocs(cmd, envs)
+		}
+	}
+
 	rootCmd.AddCommand(
 		serveCmd,
 		createCmd,
@@ -935,6 +1359,7 @@ func NewCLI() *cobra.Command {
 		pullCmd,
 		pushCmd,
 		listCmd,
+		psCmd,
 		copyCmd,
 		deleteCmd,
 	)

cmd/interactive.go

@@ -8,15 +8,17 @@ import (
 	"os"
 	"path/filepath"
 	"regexp"
+	"slices"
 	"sort"
 	"strings"
 
 	"github.com/spf13/cobra"
-	"golang.org/x/exp/slices"
 
-	"github.com/jmorganca/ollama/api"
-	"github.com/jmorganca/ollama/progress"
-	"github.com/jmorganca/ollama/readline"
+	"github.com/ollama/ollama/api"
+	"github.com/ollama/ollama/envconfig"
+	"github.com/ollama/ollama/progress"
+	"github.com/ollama/ollama/readline"
+	"github.com/ollama/ollama/types/errtypes"
 )
 
 type MultilineState int
@@ -25,64 +27,43 @@ const (
 	MultilineNone MultilineState = iota
 	MultilinePrompt
 	MultilineSystem
-	MultilineTemplate
 )
 
 func loadModel(cmd *cobra.Command, opts *runOptions) error {
-	client, err := api.ClientFromEnvironment()
-	if err != nil {
-		return err
-	}
-
 	p := progress.NewProgress(os.Stderr)
 	defer p.StopAndClear()
 
 	spinner := progress.NewSpinner("")
 	p.Add("", spinner)
 
-	showReq := api.ShowRequest{Name: opts.Model}
-	showResp, err := client.Show(cmd.Context(), &showReq)
+	client, err := api.ClientFromEnvironment()
 	if err != nil {
 		return err
 	}
-	opts.MultiModal = slices.Contains(showResp.Details.Families, "clip")
-	opts.ParentModel = showResp.Details.ParentModel
-
-	if len(showResp.Messages) > 0 {
-		opts.Messages = append(opts.Messages, showResp.Messages...)
-	}
 
 	chatReq := &api.ChatRequest{
-		Model:    opts.Model,
-		Messages: []api.Message{},
+		Model:     opts.Model,
+		KeepAlive: opts.KeepAlive,
 	}
-	err = client.Chat(cmd.Context(), chatReq, func(resp api.ChatResponse) error {
+
+	return client.Chat(cmd.Context(), chatReq, func(resp api.ChatResponse) error {
 		p.StopAndClear()
-		if len(opts.Messages) > 0 {
-			for _, msg := range opts.Messages {
-				switch msg.Role {
-				case "user":
-					fmt.Printf(">>> %s\n", msg.Content)
-				case "assistant":
-					state := &displayResponseState{}
-					displayResponse(msg.Content, opts.WordWrap, state)
-					fmt.Println()
-					fmt.Println()
-				}
+		for _, msg := range opts.Messages {
+			switch msg.Role {
+			case "user":
+				fmt.Printf(">>> %s\n", msg.Content)
+			case "assistant":
+				state := &displayResponseState{}
+				displayResponse(msg.Content, opts.WordWrap, state)
+				fmt.Println()
+				fmt.Println()
 			}
 		}
 		return nil
 	})
-	if err != nil {
-		return err
-	}
-
-	return nil
 }
 
 func generateInteractive(cmd *cobra.Command, opts runOptions) error {
-	opts.Messages = make([]api.Message, 0)
-
 	err := loadModel(cmd, &opts)
 	if err != nil {
 		return err
@@ -94,6 +75,7 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 		fmt.Fprintln(os.Stderr, "  /show           Show model information")
 		fmt.Fprintln(os.Stderr, "  /load <model>   Load a session or model")
 		fmt.Fprintln(os.Stderr, "  /save <model>   Save your current session")
+		fmt.Fprintln(os.Stderr, "  /clear          Clear session context")
 		fmt.Fprintln(os.Stderr, "  /bye            Exit")
 		fmt.Fprintln(os.Stderr, "  /?, /help       Help for a command")
 		fmt.Fprintln(os.Stderr, "  /? shortcuts    Help for keyboard shortcuts")
@@ -111,7 +93,6 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 		fmt.Fprintln(os.Stderr, "Available Commands:")
 		fmt.Fprintln(os.Stderr, "  /set parameter ...     Set a parameter")
 		fmt.Fprintln(os.Stderr, "  /set system <string>   Set system message")
-		fmt.Fprintln(os.Stderr, "  /set template <string> Set prompt template")
 		fmt.Fprintln(os.Stderr, "  /set history           Enable history")
 		fmt.Fprintln(os.Stderr, "  /set nohistory         Disable history")
 		fmt.Fprintln(os.Stderr, "  /set wordwrap          Enable wordwrap")
@@ -131,6 +112,7 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 		fmt.Fprintln(os.Stderr, "   Alt + f            Move forward (right) one word")
 		fmt.Fprintln(os.Stderr, "  Ctrl + k            Delete the sentence after the cursor")
 		fmt.Fprintln(os.Stderr, "  Ctrl + u            Delete the sentence before the cursor")
+		fmt.Fprintln(os.Stderr, "  Ctrl + w            Delete the word before the cursor")
 		fmt.Fprintln(os.Stderr, "")
 		fmt.Fprintln(os.Stderr, "  Ctrl + l            Clear the screen")
 		fmt.Fprintln(os.Stderr, "  Ctrl + c            Stop the model from responding")
@@ -161,7 +143,7 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 		fmt.Fprintln(os.Stderr, "  /set parameter repeat_penalty <float> How strongly to penalize repetitions")
 		fmt.Fprintln(os.Stderr, "  /set parameter repeat_last_n <int>    Set how far back to look for repetitions")
 		fmt.Fprintln(os.Stderr, "  /set parameter num_gpu <int>          The number of layers to send to the GPU")
-		fmt.Fprintln(os.Stderr, "  /set parameter stop \"<string>\", ...   Set the stop parameters")
+		fmt.Fprintln(os.Stderr, "  /set parameter stop <string> <string> ...   Set the stop parameters")
 		fmt.Fprintln(os.Stderr, "")
 	}
 
@@ -175,6 +157,10 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 		return err
 	}
 
+	if envconfig.NoHistory {
+		scanner.HistoryDisable()
+	}
+
 	fmt.Print(readline.StartBracketedPaste)
 	defer fmt.Printf(readline.EndBracketedPaste)
 
@@ -216,10 +202,6 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 				opts.Messages = append(opts.Messages, api.Message{Role: "system", Content: opts.System})
 				fmt.Println("Set system message.")
 				sb.Reset()
-			case MultilineTemplate:
-				opts.Template = sb.String()
-				fmt.Println("Set prompt template.")
-				sb.Reset()
 			}
 
 			multiline = MultilineNone
@@ -275,11 +257,22 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 			fn := func(resp api.ProgressResponse) error { return nil }
 			err = client.Create(cmd.Context(), req, fn)
 			if err != nil {
-				fmt.Println("error: couldn't save model")
+				if strings.Contains(err.Error(), errtypes.InvalidModelNameErrMsg) {
+					fmt.Printf("error: The model name '%s' is invalid\n", args[1])
+					continue
+				}
 				return err
 			}
 			fmt.Printf("Created new model '%s'\n", args[1])
 			continue
+		case strings.HasPrefix(line, "/clear"):
+			opts.Messages = []api.Message{}
+			if opts.System != "" {
+				newMessage := api.Message{Role: "system", Content: opts.System}
+				opts.Messages = append(opts.Messages, newMessage)
+			}
+			fmt.Println("Cleared session context")
+			continue
 		case strings.HasPrefix(line, "/set"):
 			args := strings.Fields(line)
 			if len(args) > 1 {
@@ -295,10 +288,14 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 					opts.WordWrap = false
 					fmt.Println("Set 'nowordwrap' mode.")
 				case "verbose":
-					cmd.Flags().Set("verbose", "true")
+					if err := cmd.Flags().Set("verbose", "true"); err != nil {
+						return err
+					}
 					fmt.Println("Set 'verbose' mode.")
 				case "quiet":
-					cmd.Flags().Set("verbose", "false")
+					if err := cmd.Flags().Set("verbose", "false"); err != nil {
+						return err
+					}
 					fmt.Println("Set 'quiet' mode.")
 				case "format":
 					if len(args) < 3 || args[2] != "json" {
@@ -323,17 +320,13 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 					}
 					fmt.Printf("Set parameter '%s' to '%s'\n", args[2], strings.Join(params, ", "))
 					opts.Options[args[2]] = fp[args[2]]
-				case "system", "template":
+				case "system":
 					if len(args) < 3 {
 						usageSet()
 						continue
 					}
 
-					if args[1] == "system" {
-						multiline = MultilineSystem
-					} else if args[1] == "template" {
-						multiline = MultilineTemplate
-					}
+					multiline = MultilineSystem
 
 					line := strings.Join(args[2:], " ")
 					line, ok := strings.CutPrefix(line, `"""`)
@@ -353,23 +346,17 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 						continue
 					}
 
-					if args[1] == "system" {
-						opts.System = sb.String() // for display in modelfile
-						newMessage := api.Message{Role: "system", Content: sb.String()}
-						// Check if the slice is not empty and the last message is from 'system'
-						if len(opts.Messages) > 0 && opts.Messages[len(opts.Messages)-1].Role == "system" {
-							// Replace the last message
-							opts.Messages[len(opts.Messages)-1] = newMessage
-						} else {
-							opts.Messages = append(opts.Messages, newMessage)
-						}
-						fmt.Println("Set system message.")
-						sb.Reset()
-					} else if args[1] == "template" {
-						opts.Template = sb.String()
-						fmt.Println("Set prompt template.")
-						sb.Reset()
+					opts.System = sb.String() // for display in modelfile
+					newMessage := api.Message{Role: "system", Content: sb.String()}
+					// Check if the slice is not empty and the last message is from 'system'
+					if len(opts.Messages) > 0 && opts.Messages[len(opts.Messages)-1].Role == "system" {
+						// Replace the last message
+						opts.Messages[len(opts.Messages)-1] = newMessage
+					} else {
+						opts.Messages = append(opts.Messages, newMessage)
 					}
+					fmt.Println("Set system message.")
+					sb.Reset()
 
 					sb.Reset()
 					continue
@@ -390,7 +377,6 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 				req := &api.ShowRequest{
 					Name:     opts.Model,
 					System:   opts.System,
-					Template: opts.Template,
 					Options:  opts.Options,
 				}
 				resp, err := client.Show(cmd.Context(), req)
@@ -401,15 +387,7 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 
 				switch args[1] {
 				case "info":
-					fmt.Println("Model details:")
-					if len(resp.Details.Families) > 0 {
-						fmt.Printf("Family              %s\n", strings.Join(resp.Details.Families, ", "))
-					} else if resp.Details.Family != "" {
-						fmt.Printf("Family              %s\n", resp.Details.Family)
-					}
-					fmt.Printf("Parameter Size      %s\n", resp.Details.ParameterSize)
-					fmt.Printf("Quantization Level  %s\n", resp.Details.QuantizationLevel)
-					fmt.Println("")
+					showInfo(resp)
 				case "license":
 					if resp.License == "" {
 						fmt.Println("No license was specified for this model.")
@@ -442,12 +420,9 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 						fmt.Println("No system message was specified for this model.")
 					}
 				case "template":
-					switch {
-					case opts.Template != "":
-						fmt.Println(opts.Template + "\n")
-					case resp.Template != "":
+					if resp.Template != "" {
 						fmt.Println(resp.Template)
-					default:
+					} else {
 						fmt.Println("No prompt template was specified for this model.")
 					}
 				default:
@@ -541,10 +516,6 @@ func buildModelfile(opts runOptions) string {
 		fmt.Fprintf(&mf, "SYSTEM \"\"\"%s\"\"\"\n", opts.System)
 	}
 
-	if opts.Template != "" {
-		fmt.Fprintf(&mf, "TEMPLATE \"\"\"%s\"\"\"\n", opts.Template)
-	}
-
 	keys := make([]string, 0)
 	for k := range opts.Options {
 		keys = append(keys, k)

cmd/interactive_test.go

@@ -6,8 +6,9 @@ import (
 	"text/template"
 
 	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 
-	"github.com/jmorganca/ollama/api"
+	"github.com/ollama/ollama/api"
 )
 
 func TestExtractFilenames(t *testing.T) {
@@ -58,7 +59,6 @@ func TestModelfileBuilder(t *testing.T) {
 	opts := runOptions{
 		Model:    "hork",
 		System:   "You are part horse and part shark, but all hork. Do horklike things",
-		Template: "This is a template.",
 		Messages: []api.Message{
 			{Role: "user", Content: "Hey there hork!"},
 			{Role: "assistant", Content: "Yes it is true, I am half horse, half shark."},
@@ -74,7 +74,6 @@ func TestModelfileBuilder(t *testing.T) {
 	mf := buildModelfile(opts)
 	expectedModelfile := `FROM {{.Model}}
 SYSTEM """{{.System}}"""
-TEMPLATE """{{.Template}}"""
 PARAMETER penalize_newline false
 PARAMETER seed 42
 PARAMETER stop [hi there]
@@ -85,18 +84,17 @@ MESSAGE assistant """Yes it is true, I am half horse, half shark."""
 `
 
 	tmpl, err := template.New("").Parse(expectedModelfile)
-	assert.Nil(t, err)
+	require.NoError(t, err)
 
 	var buf bytes.Buffer
 	err = tmpl.Execute(&buf, opts)
-	assert.Nil(t, err)
+	require.NoError(t, err)
 	assert.Equal(t, buf.String(), mf)
 
 	opts.ParentModel = "horseshark"
 	mf = buildModelfile(opts)
 	expectedModelfile = `FROM {{.ParentModel}}
 SYSTEM """{{.System}}"""
-TEMPLATE """{{.Template}}"""
 PARAMETER penalize_newline false
 PARAMETER seed 42
 PARAMETER stop [hi there]
@@ -107,10 +105,10 @@ MESSAGE assistant """Yes it is true, I am half horse, half shark."""
 `
 
 	tmpl, err = template.New("").Parse(expectedModelfile)
-	assert.Nil(t, err)
+	require.NoError(t, err)
 
 	var parentBuf bytes.Buffer
 	err = tmpl.Execute(&parentBuf, opts)
-	assert.Nil(t, err)
+	require.NoError(t, err)
 	assert.Equal(t, parentBuf.String(), mf)
 }

cmd/start.go

@@ -0,0 +1,27 @@
+//go:build darwin || windows
+
+package cmd
+
+import (
+	"context"
+	"errors"
+	"time"
+
+	"github.com/ollama/ollama/api"
+)
+
+func waitForServer(ctx context.Context, client *api.Client) error {
+	// wait for the server to start
+	timeout := time.After(5 * time.Second)
+	tick := time.Tick(500 * time.Millisecond)
+	for {
+		select {
+		case <-timeout:
+			return errors.New("timed out waiting for server to start")
+		case <-tick:
+			if err := client.Heartbeat(ctx); err == nil {
+				return nil // server has started
+			}
+		}
+	}
+}

cmd/start_darwin.go

@@ -7,7 +7,7 @@ import (
 	"os/exec"
 	"strings"
 
-	"github.com/jmorganca/ollama/api"
+	"github.com/ollama/ollama/api"
 )
 
 func startApp(ctx context.Context, client *api.Client) error {

cmd/start_default.go

@@ -6,7 +6,7 @@ import (
 	"context"
 	"fmt"
 
-	"github.com/jmorganca/ollama/api"
+	"github.com/ollama/ollama/api"
 )
 
 func startApp(ctx context.Context, client *api.Client) error {

cmd/start_windows.go

@@ -10,7 +10,7 @@ import (
 	"strings"
 	"syscall"
 
-	"github.com/jmorganca/ollama/api"
+	"github.com/ollama/ollama/api"
 )
 
 func startApp(ctx context.Context, client *api.Client) error {

convert/convert.go

@@ -0,0 +1,200 @@
+package convert
+
+import (
+	"cmp"
+	"encoding/binary"
+	"encoding/json"
+	"fmt"
+	"io"
+	"log/slog"
+	"os"
+	"path/filepath"
+	"slices"
+	"strings"
+
+	"google.golang.org/protobuf/proto"
+
+	"github.com/ollama/ollama/convert/sentencepiece"
+	"github.com/ollama/ollama/llm"
+)
+
+const (
+	_ int32 = iota
+	tokenTypeNormal
+	tokenTypeUnknown
+	tokenTypeControl
+	tokenTypeUserDefined
+	tokenTypeUnused
+	tokenTypeByte
+)
+
+type Params struct {
+	Architectures     []string `json:"architectures"`
+	VocabSize         int      `json:"vocab_size"`
+	HiddenSize        int      `json:"hidden_size"`       // n_embd
+	HiddenLayers      int      `json:"num_hidden_layers"` // n_layer
+	ContextSize       int      `json:"max_position_embeddings"`
+	IntermediateSize  int      `json:"intermediate_size"`
+	AttentionHeads    int      `json:"num_attention_heads"` // n_head
+	KeyValHeads       int      `json:"num_key_value_heads"`
+	NormEPS           float64  `json:"rms_norm_eps"`
+	BoSTokenID        int      `json:"bos_token_id"`
+	EoSTokenID        int      `json:"eos_token_id"`
+	HeadDimension     int      `json:"head_dim"`
+	PaddingTokenID    int      `json:"pad_token_id"`
+	RopeFrequencyBase float64  `json:"rope_theta"`
+
+	Experts     int `json:"num_local_experts"`
+	ExpertsUsed int `json:"num_experts_per_tok"`
+
+	PreTokenizer string
+
+	ByteOrder
+}
+
+type ByteOrder interface {
+	binary.ByteOrder
+	binary.AppendByteOrder
+}
+
+type ModelArch interface {
+	GetTensors() error
+	LoadVocab() error
+	WriteGGUF(io.WriteSeeker) error
+}
+
+type ModelFormat interface {
+	GetLayerName(string) (string, error)
+	GetTensors(string, *Params) ([]llm.Tensor, error)
+	GetParams(string) (*Params, error)
+	GetModelArch(string, string, *Params) (ModelArch, error)
+}
+
+type ModelData struct {
+	Path    string
+	Name    string
+	Params  *Params
+	Vocab   *Vocab
+	Tensors []llm.Tensor
+	Format  ModelFormat
+}
+
+func GetModelFormat(dirname string) (ModelFormat, error) {
+	files, err := filepath.Glob(filepath.Join(dirname, "*"))
+	if err != nil {
+		return nil, err
+	}
+
+	for _, fn := range files {
+		if strings.HasSuffix(fn, ".safetensors") {
+			return &SafetensorFormat{}, nil
+		} else if strings.HasSuffix(fn, ".bin") || strings.HasSuffix(fn, ".pth") {
+			slog.Debug("model is torch")
+			return &TorchFormat{}, nil
+		}
+	}
+
+	return nil, fmt.Errorf("couldn't determine model format")
+}
+
+// Details on gguf's tokenizer can be found at:
+// https://github.com/ggerganov/ggml/blob/master/docs/gguf.md#tokenizer
+type Vocab struct {
+	Tokens []string
+	Scores []float32
+	Types  []int32
+	Merges []string
+}
+
+func LoadSentencePieceTokens(dirpath string, params *Params) (*Vocab, error) {
+	slog.Info(fmt.Sprintf("reading vocab from %s", filepath.Join(dirpath, "tokenizer.model")))
+	in, err := os.ReadFile(filepath.Join(dirpath, "tokenizer.model"))
+	if err != nil {
+		return nil, err
+	}
+
+	// To regenerate sentencepiece from the protobufs use:
+	// protoc -I=./ --go_out=./ sentencepiece_model.proto
+	modelProto := &sentencepiece.ModelProto{}
+	if err := proto.Unmarshal(in, modelProto); err != nil {
+		return nil, err
+	}
+
+	v := &Vocab{
+		Tokens: make([]string, 0),
+		Scores: make([]float32, 0),
+		Types:  make([]int32, 0),
+	}
+
+	pieces := modelProto.GetPieces()
+	for _, p := range pieces {
+		v.Tokens = append(v.Tokens, p.GetPiece())
+		v.Scores = append(v.Scores, p.GetScore())
+		t := p.GetType()
+		switch t {
+		case sentencepiece.ModelProto_SentencePiece_UNKNOWN:
+		case sentencepiece.ModelProto_SentencePiece_CONTROL:
+		case sentencepiece.ModelProto_SentencePiece_UNUSED:
+		case sentencepiece.ModelProto_SentencePiece_BYTE:
+		default:
+			t = sentencepiece.ModelProto_SentencePiece_NORMAL
+		}
+		v.Types = append(v.Types, int32(t))
+	}
+
+	slog.Info(fmt.Sprintf("vocab size: %d", len(v.Tokens)))
+
+	// add any additional tokens
+	addIn, err := os.ReadFile(filepath.Join(dirpath, "added_tokens.json"))
+	if os.IsNotExist(err) {
+		return v, nil
+	} else if err != nil {
+		return nil, err
+	}
+
+	slog.Info("reading user defined tokens")
+
+	var extraTokenData map[string]int
+	if err := json.Unmarshal(addIn, &extraTokenData); err != nil {
+		return nil, err
+	}
+
+	type token struct {
+		key string
+		pos int
+	}
+
+	extraTokens := make([]token, 0)
+	for k, id := range extraTokenData {
+		extraTokens = append(extraTokens, token{k, id})
+	}
+
+	slices.SortFunc(extraTokens, func(a, b token) int {
+		return cmp.Compare(a.pos, b.pos)
+	})
+
+	numToks := len(v.Tokens)
+
+	for cnt, t := range extraTokens {
+		// the token id should match the specific index for the total number of tokens
+		if t.pos != cnt+numToks {
+			return nil, fmt.Errorf("token ID '%d' for '%s' doesn't match total token size", t.pos, t.key)
+		}
+		v.Tokens = append(v.Tokens, t.key)
+		v.Scores = append(v.Scores, -1000.0)
+		v.Types = append(v.Types, tokenTypeUserDefined)
+	}
+	slog.Info(fmt.Sprintf("vocab size w/ extra tokens: %d", len(v.Tokens)))
+
+	if params.VocabSize > len(v.Tokens) {
+		missingTokens := params.VocabSize - len(v.Tokens)
+		slog.Warn(fmt.Sprintf("vocab is missing %d tokens", missingTokens))
+		for cnt := range missingTokens {
+			v.Tokens = append(v.Tokens, fmt.Sprintf("<dummy%05d>", cnt+1))
+			v.Scores = append(v.Scores, -1)
+			v.Types = append(v.Types, tokenTypeUserDefined)
+		}
+	}
+
+	return v, nil
+}

convert/convert_test.go

@@ -0,0 +1,103 @@
+//go:build slow
+
+package convert
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/ollama/ollama/llm"
+)
+
+func convertFull(t *testing.T, p string) (llm.KV, llm.Tensors) {
+	t.Helper()
+
+	mf, err := GetModelFormat(p)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	params, err := mf.GetParams(p)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	arch, err := mf.GetModelArch("", p, params)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	if err := arch.LoadVocab(); err != nil {
+		t.Fatal(err)
+	}
+
+	if err := arch.GetTensors(); err != nil {
+		t.Fatal(err)
+	}
+
+	f, err := os.CreateTemp(t.TempDir(), "f16")
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer f.Close()
+
+	if err := arch.WriteGGUF(f); err != nil {
+		t.Fatal(err)
+	}
+
+	r, err := os.Open(f.Name())
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer r.Close()
+
+	m, _, err := llm.DecodeGGML(r)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	return m.KV(), m.Tensors()
+}
+
+func TestConvertFull(t *testing.T) {
+	cases := []struct {
+		path    string
+		arch    string
+		tensors int
+		layers  int
+	}{
+		{"Meta-Llama-3-8B-Instruct", "llama", 291, 35},
+		{"Mistral-7B-Instruct-v0.2", "llama", 291, 35},
+		{"Mixtral-8x7B-Instruct-v0.1", "llama", 291, 35},
+		{"gemma-2b-it", "gemma", 164, 20},
+	}
+
+	for _, tt := range cases {
+		t.Run(tt.path, func(t *testing.T) {
+			p := filepath.Join("testdata", tt.path)
+			if _, err := os.Stat(p); err != nil {
+				t.Skipf("%s not found", p)
+			}
+
+			kv, tensors := convertFull(t, p)
+
+			if kv.Architecture() != tt.arch {
+				t.Fatalf("expected llama, got %s", kv.Architecture())
+			}
+
+			if kv.FileType().String() != "F16" {
+				t.Fatalf("expected F16, got %s", kv.FileType())
+			}
+
+			if len(tensors) != tt.tensors {
+				t.Fatalf("expected %d tensors, got %d", tt.tensors, len(tensors))
+			}
+
+			layers := tensors.Layers()
+			if len(layers) != tt.layers {
+				t.Fatalf("expected %d layers, got %d", tt.layers, len(layers))
+			}
+		})
+	}
+}

convert/gemma.go

@@ -0,0 +1,102 @@
+package convert
+
+import (
+	"fmt"
+	"io"
+	"log/slog"
+	"strings"
+
+	"github.com/pdevine/tensor"
+	"github.com/pdevine/tensor/native"
+
+	"github.com/ollama/ollama/llm"
+)
+
+type GemmaModel struct {
+	ModelData
+}
+
+func addOnes(data []float32, vectorSize int) ([]float32, error) {
+	n := tensor.New(tensor.WithShape(vectorSize), tensor.WithBacking(data))
+	ones := tensor.Ones(tensor.Float32, vectorSize)
+
+	n, err := n.Add(ones)
+	if err != nil {
+		return nil, err
+	}
+
+	ts, err := native.SelectF32(n, 0)
+	if err != nil {
+		return nil, err
+	}
+
+	var f32s []float32
+	for _, t := range ts {
+		f32s = append(f32s, t...)
+	}
+
+	return f32s, nil
+}
+
+func (m *GemmaModel) GetTensors() error {
+	t, err := m.Format.GetTensors(m.Path, m.Params)
+	if err != nil {
+		return err
+	}
+
+	slog.Debug(fmt.Sprintf("Total tensors: %d", len(t)))
+	for _, l := range t {
+		if strings.HasSuffix(l.Name, "norm.weight") {
+			wt := l.WriterTo.(safetensorWriterTo)
+			wt.repacker = m.Repack
+			l.WriterTo = wt
+		}
+		m.Tensors = append(m.Tensors, l)
+	}
+
+	return nil
+}
+
+func (m *GemmaModel) LoadVocab() error {
+	v, err := LoadSentencePieceTokens(m.Path, m.Params)
+	if err != nil {
+		return err
+	}
+	m.Vocab = v
+	return nil
+}
+
+func (m *GemmaModel) Repack(_ string, data []float32, shape []uint64) ([]float32, error) {
+	return addOnes(data, int(shape[0]))
+}
+
+func (m *GemmaModel) WriteGGUF(ws io.WriteSeeker) error {
+	kv := llm.KV{
+		"general.architecture":                   "gemma",
+		"general.name":                           m.Name,
+		"gemma.context_length":                   uint32(m.Params.ContextSize),
+		"gemma.embedding_length":                 uint32(m.Params.HiddenSize),
+		"gemma.block_count":                      uint32(m.Params.HiddenLayers),
+		"gemma.feed_forward_length":              uint32(m.Params.IntermediateSize),
+		"gemma.attention.head_count":             uint32(m.Params.AttentionHeads),
+		"gemma.attention.head_count_kv":          uint32(m.Params.KeyValHeads),
+		"gemma.attention.layer_norm_rms_epsilon": float32(m.Params.NormEPS),
+		"gemma.attention.key_length":             uint32(m.Params.HeadDimension),
+		"gemma.attention.value_length":           uint32(m.Params.HeadDimension),
+		"general.file_type":                      uint32(1),
+		"tokenizer.ggml.model":                   "llama",
+
+		"tokenizer.ggml.tokens":     m.Vocab.Tokens,
+		"tokenizer.ggml.scores":     m.Vocab.Scores,
+		"tokenizer.ggml.token_type": m.Vocab.Types,
+
+		"tokenizer.ggml.bos_token_id":     uint32(m.Params.BoSTokenID),
+		"tokenizer.ggml.eos_token_id":     uint32(m.Params.EoSTokenID),
+		"tokenizer.ggml.padding_token_id": uint32(m.Params.PaddingTokenID),
+		"tokenizer.ggml.unknown_token_id": uint32(3),
+		"tokenizer.ggml.add_bos_token":    true,
+		"tokenizer.ggml.add_eos_token":    false,
+	}
+
+	return llm.NewGGUFV3(m.Params.ByteOrder).Encode(ws, kv, m.Tensors)
+}

convert/llama.go

@@ -0,0 +1,159 @@
+package convert
+
+import (
+	"cmp"
+	"errors"
+	"fmt"
+	"io"
+	"os"
+	"path/filepath"
+	"regexp"
+	"strings"
+
+	"github.com/pdevine/tensor"
+	"github.com/pdevine/tensor/native"
+
+	"github.com/ollama/ollama/llm"
+)
+
+type LlamaModel struct {
+	ModelData
+}
+
+func (m *LlamaModel) GetTensors() error {
+	t, err := m.Format.GetTensors(m.Path, m.Params)
+	if err != nil {
+		return err
+	}
+
+	pattern := `^blk\.[0-9]+\.attn_(?P<layer>q|k)\.weight$`
+	re, err := regexp.Compile(pattern)
+	if err != nil {
+		return err
+	}
+
+	for _, l := range t {
+		matches := re.FindAllStringSubmatch(l.Name, -1)
+		if len(matches) > 0 {
+			switch m.Format.(type) {
+			case *TorchFormat:
+				wt := l.WriterTo.(torchWriterTo)
+				wt.repacker = m.Repack
+				l.WriterTo = wt
+			case *SafetensorFormat:
+				wt := l.WriterTo.(safetensorWriterTo)
+				wt.repacker = m.Repack
+				l.WriterTo = wt
+			}
+		}
+		m.Tensors = append(m.Tensors, l)
+	}
+
+	return nil
+}
+
+func (m *LlamaModel) LoadVocab() (err error) {
+	pre, ts, merges, err := parseTokens(filepath.Join(m.Path, "tokenizer.json"))
+	if errors.Is(err, os.ErrNotExist) {
+		return nil
+	} else if err != nil {
+		return err
+	}
+
+	m.Vocab = &Vocab{}
+	for _, t := range ts {
+		m.Vocab.Tokens = append(m.Vocab.Tokens, t.Content)
+		m.Vocab.Types = append(m.Vocab.Types, t.Type())
+	}
+
+	m.Vocab.Merges = merges
+	m.Params.PreTokenizer = pre
+	return nil
+}
+
+func (m *LlamaModel) WriteGGUF(ws io.WriteSeeker) error {
+	kv := llm.KV{
+		"general.architecture":                   "llama",
+		"general.name":                           m.Name,
+		"llama.vocab_size":                       uint32(len(m.Vocab.Tokens)),
+		"llama.context_length":                   uint32(m.Params.ContextSize),
+		"llama.embedding_length":                 uint32(m.Params.HiddenSize),
+		"llama.block_count":                      uint32(m.Params.HiddenLayers),
+		"llama.feed_forward_length":              uint32(m.Params.IntermediateSize),
+		"llama.rope.freq_base":                   float32(m.Params.RopeFrequencyBase),
+		"llama.rope.dimension_count":             uint32(m.Params.HiddenSize / m.Params.AttentionHeads),
+		"llama.attention.head_count":             uint32(m.Params.AttentionHeads),
+		"llama.attention.head_count_kv":          uint32(m.Params.KeyValHeads),
+		"llama.attention.layer_norm_rms_epsilon": float32(m.Params.NormEPS),
+		"general.file_type":                      uint32(1),
+		"tokenizer.ggml.model":                   "gpt2",
+
+		"tokenizer.ggml.pre":        m.Params.PreTokenizer,
+		"tokenizer.ggml.tokens":     m.Vocab.Tokens,
+		"tokenizer.ggml.token_type": m.Vocab.Types,
+
+		"tokenizer.ggml.bos_token_id":     uint32(m.Params.BoSTokenID),
+		"tokenizer.ggml.eos_token_id":     uint32(m.Params.EoSTokenID),
+		"tokenizer.ggml.unknown_token_id": uint32(0),
+	}
+
+	if len(m.Vocab.Merges) > 0 {
+		kv["tokenizer.ggml.merges"] = m.Vocab.Merges
+	} else {
+		kv["tokenizer.ggml.scores"] = m.Vocab.Scores
+	}
+
+	return llm.NewGGUFV3(m.Params.ByteOrder).Encode(ws, kv, m.Tensors)
+}
+
+func (m *LlamaModel) Repack(name string, data []float32, shape []uint64) ([]float32, error) {
+	return llamaRepack(name, m.Params, data, shape)
+}
+
+func llamaRepack(name string, params *Params, data []float32, shape []uint64) ([]float32, error) {
+	var dims []int
+	for _, dim := range shape {
+		if dim != 0 {
+			dims = append(dims, int(dim))
+		}
+	}
+
+	var heads int
+	switch {
+	case strings.HasSuffix(name, "attn_q.weight"):
+		heads = params.AttentionHeads
+	case strings.HasSuffix(name, "attn_k.weight"):
+		heads = cmp.Or(params.KeyValHeads, params.AttentionHeads)
+	default:
+		return nil, fmt.Errorf("unknown tensor name: %s", name)
+	}
+
+	n := tensor.New(tensor.WithShape(dims...), tensor.WithBacking(data))
+	if err := n.Reshape(append([]int{heads, 2, dims[0] / heads / 2}, dims[1:]...)...); err != nil {
+		return nil, err
+	}
+
+	if err := n.T(0, 2, 1, 3); err != nil {
+		return nil, err
+	}
+
+	if err := n.Reshape(dims...); err != nil {
+		return nil, err
+	}
+
+	if err := n.Transpose(); err != nil {
+		return nil, err
+	}
+
+	ts, err := native.SelectF32(n, 1)
+	if err != nil {
+		return nil, err
+	}
+
+	var f32s []float32
+	for _, t := range ts {
+		f32s = append(f32s, t...)
+	}
+
+	return f32s, nil
+}

convert/mistral.go

@@ -0,0 +1,79 @@
+package convert
+
+import (
+	"io"
+	"regexp"
+
+	"github.com/ollama/ollama/llm"
+)
+
+type MistralModel struct {
+	ModelData
+}
+
+func (m *MistralModel) GetTensors() error {
+	t, err := m.Format.GetTensors(m.Path, m.Params)
+	if err != nil {
+		return err
+	}
+
+	pattern := `^blk\.[0-9]+\.attn_(?P<layer>q|k)\.weight$`
+	re, err := regexp.Compile(pattern)
+	if err != nil {
+		return err
+	}
+
+	for _, l := range t {
+		matches := re.FindAllStringSubmatch(l.Name, -1)
+		if len(matches) > 0 {
+			wt := l.WriterTo.(safetensorWriterTo)
+			wt.repacker = m.Repack
+			l.WriterTo = wt
+		}
+		m.Tensors = append(m.Tensors, l)
+	}
+
+	return nil
+}
+
+func (m *MistralModel) LoadVocab() error {
+	v, err := LoadSentencePieceTokens(m.Path, m.Params)
+	if err != nil {
+		return err
+	}
+	m.Vocab = v
+	return nil
+}
+
+func (m *MistralModel) WriteGGUF(ws io.WriteSeeker) error {
+	kv := llm.KV{
+		"general.architecture":                   "llama",
+		"general.name":                           m.Name,
+		"llama.context_length":                   uint32(m.Params.ContextSize),
+		"llama.embedding_length":                 uint32(m.Params.HiddenSize),
+		"llama.block_count":                      uint32(m.Params.HiddenLayers),
+		"llama.feed_forward_length":              uint32(m.Params.IntermediateSize),
+		"llama.rope.dimension_count":             uint32(m.Params.HiddenSize / m.Params.AttentionHeads),
+		"llama.attention.head_count":             uint32(m.Params.AttentionHeads),
+		"llama.attention.head_count_kv":          uint32(m.Params.KeyValHeads),
+		"llama.attention.layer_norm_rms_epsilon": float32(m.Params.NormEPS),
+		"general.file_type":                      uint32(1),
+		"tokenizer.ggml.model":                   "llama",
+
+		"tokenizer.ggml.tokens":     m.Vocab.Tokens,
+		"tokenizer.ggml.scores":     m.Vocab.Scores,
+		"tokenizer.ggml.token_type": m.Vocab.Types,
+
+		"tokenizer.ggml.bos_token_id":     uint32(m.Params.BoSTokenID),
+		"tokenizer.ggml.eos_token_id":     uint32(m.Params.EoSTokenID),
+		"tokenizer.ggml.add_bos_token":    true,
+		"tokenizer.ggml.add_eos_token":    false,
+		"tokenizer.ggml.unknown_token_id": uint32(0),
+	}
+
+	return llm.NewGGUFV3(m.Params.ByteOrder).Encode(ws, kv, m.Tensors)
+}
+
+func (m *MistralModel) Repack(name string, data []float32, shape []uint64) ([]float32, error) {
+	return llamaRepack(name, m.Params, data, shape)
+}

convert/mixtral.go

@@ -0,0 +1,87 @@
+package convert
+
+import (
+	"io"
+	"regexp"
+
+	"github.com/ollama/ollama/llm"
+)
+
+type MixtralModel struct {
+	ModelData
+}
+
+func (m *MixtralModel) GetTensors() error {
+	t, err := m.Format.GetTensors(m.Path, m.Params)
+	if err != nil {
+		return err
+	}
+
+	pattern := `^blk\.[0-9]+\.attn_(?P<layer>q|k)\.weight$`
+	re, err := regexp.Compile(pattern)
+	if err != nil {
+		return err
+	}
+
+	for _, l := range t {
+		matches := re.FindAllStringSubmatch(l.Name, -1)
+		if len(matches) > 0 {
+			wt := l.WriterTo.(safetensorWriterTo)
+			wt.repacker = m.Repack
+			l.WriterTo = wt
+		}
+		m.Tensors = append(m.Tensors, l)
+	}
+
+	return nil
+}
+
+func (m *MixtralModel) LoadVocab() error {
+	v, err := LoadSentencePieceTokens(m.Path, m.Params)
+	if err != nil {
+		return err
+	}
+	m.Vocab = v
+	return nil
+}
+
+func (m *MixtralModel) WriteGGUF(ws io.WriteSeeker) error {
+	kv := llm.KV{
+		"general.architecture":          "llama",
+		"general.name":                  m.Name,
+		"llama.block_count":             uint32(m.Params.HiddenLayers),
+		"llama.context_length":          uint32(m.Params.ContextSize),
+		"llama.embedding_length":        uint32(m.Params.HiddenSize),
+		"llama.feed_forward_length":     uint32(m.Params.IntermediateSize),
+		"llama.attention.head_count":    uint32(m.Params.AttentionHeads),
+		"llama.attention.head_count_kv": uint32(m.Params.KeyValHeads),
+
+		"llama.rope.freq_base":                   float32(m.Params.RopeFrequencyBase),
+		"llama.attention.layer_norm_rms_epsilon": float32(m.Params.NormEPS),
+
+		"llama.expert_count":      uint32(m.Params.Experts),
+		"llama.expert_used_count": uint32(m.Params.ExpertsUsed),
+
+		"llama.vocab_size":           uint32(len(m.Vocab.Tokens)),
+		"llama.rope.dimension_count": uint32(m.Params.HiddenSize / m.Params.AttentionHeads),
+
+		"general.file_type":    uint32(1),
+		"tokenizer.ggml.model": "llama",
+
+		"tokenizer.ggml.tokens":     m.Vocab.Tokens,
+		"tokenizer.ggml.scores":     m.Vocab.Scores,
+		"tokenizer.ggml.token_type": m.Vocab.Types,
+
+		"tokenizer.ggml.bos_token_id":     uint32(m.Params.BoSTokenID),
+		"tokenizer.ggml.eos_token_id":     uint32(m.Params.EoSTokenID),
+		"tokenizer.ggml.unknown_token_id": uint32(0),
+		"tokenizer.ggml.add_bos_token":    true,
+		"tokenizer.ggml.add_eos_token":    false,
+	}
+
+	return llm.NewGGUFV3(m.Params.ByteOrder).Encode(ws, kv, m.Tensors)
+}
+
+func (m *MixtralModel) Repack(name string, data []float32, shape []uint64) ([]float32, error) {
+	return llamaRepack(name, m.Params, data, shape)
+}

convert/safetensors.go

@@ -0,0 +1,309 @@
+package convert
+
+import (
+	"bytes"
+	"encoding/binary"
+	"encoding/json"
+	"fmt"
+	"io"
+	"os"
+	"path/filepath"
+	"regexp"
+	"slices"
+	"strings"
+
+	"github.com/d4l3k/go-bfloat16"
+	"github.com/x448/float16"
+
+	"github.com/ollama/ollama/llm"
+)
+
+type safetensorWriterTo struct {
+	t *llm.Tensor
+
+	params *Params
+	bo     ByteOrder
+
+	filename string
+	dtype    string
+
+	offset, size int64
+	repacker     func(string, []float32, []uint64) ([]float32, error)
+}
+
+type safetensorMetadata struct {
+	Type    string   `json:"dtype"`
+	Shape   []uint64 `json:"shape"`
+	Offsets []int64  `json:"data_offsets"`
+}
+
+type SafetensorFormat struct{}
+
+func (m *SafetensorFormat) GetTensors(dirpath string, params *Params) ([]llm.Tensor, error) {
+	var tensors []llm.Tensor
+	matches, err := filepath.Glob(filepath.Join(dirpath, "*.safetensors"))
+	if err != nil {
+		return nil, err
+	}
+
+	var offset uint64
+	for _, f := range matches {
+		var t []llm.Tensor
+		var err error
+		t, offset, err = m.readTensors(f, offset, params)
+		if err != nil {
+			return nil, err
+		}
+
+		tensors = append(tensors, t...)
+	}
+	return tensors, nil
+}
+
+func (m *SafetensorFormat) readTensors(fn string, offset uint64, params *Params) ([]llm.Tensor, uint64, error) {
+	f, err := os.Open(fn)
+	if err != nil {
+		return nil, 0, err
+	}
+	defer f.Close()
+
+	var n int64
+	if err := binary.Read(f, binary.LittleEndian, &n); err != nil {
+		return nil, 0, err
+	}
+
+	b := bytes.NewBuffer(make([]byte, 0, n))
+	if _, err = io.CopyN(b, f, n); err != nil {
+		return nil, 0, err
+	}
+
+	var headers map[string]safetensorMetadata
+	if err := json.NewDecoder(b).Decode(&headers); err != nil {
+		return nil, 0, err
+	}
+
+	var keys []string
+	for key := range headers {
+		if !strings.HasSuffix(key, "self_attn.rotary_embd.inv_freq") {
+			keys = append(keys, key)
+		}
+	}
+
+	slices.Sort(keys)
+
+	var tensors []llm.Tensor
+	for _, key := range keys {
+		value := headers[key]
+
+		var kind uint32
+		switch len(value.Shape) {
+		case 0:
+			// valuedata
+			continue
+		case 2:
+			kind = 1
+		}
+
+		name, err := m.GetLayerName(key)
+		if err != nil {
+			return nil, 0, err
+		}
+
+		shape := make([]uint64, len(value.Shape))
+		copy(shape, value.Shape)
+
+		pad := func(s int64) int64 {
+			return 8 + n + s
+		}
+
+		t := llm.Tensor{
+			Name:   name,
+			Kind:   kind,
+			Offset: offset,
+			Shape:  shape,
+		}
+
+		t.WriterTo = safetensorWriterTo{
+			t:        &t,
+			params:   params,
+			bo:       params.ByteOrder,
+			filename: fn,
+			dtype:    value.Type,
+			offset:   pad(value.Offsets[0]),
+			size:     pad(value.Offsets[1]) - pad(value.Offsets[0]),
+		}
+
+		offset += t.Size()
+		tensors = append(tensors, t)
+	}
+
+	return tensors, offset, nil
+}
+
+func (m *SafetensorFormat) GetParams(dirpath string) (*Params, error) {
+	f, err := os.Open(filepath.Join(dirpath, "config.json"))
+	if err != nil {
+		return nil, err
+	}
+	defer f.Close()
+
+	var params Params
+
+	if err := json.NewDecoder(f).Decode(&params); err != nil {
+		return nil, err
+	}
+
+	params.ByteOrder = binary.LittleEndian
+	return &params, nil
+}
+
+func (m *SafetensorFormat) GetLayerName(n string) (string, error) {
+	directMap := map[string]string{
+		"model.embed_tokens.weight": "token_embd.weight",
+		"lm_head.weight":            "output.weight",
+		"model.norm.weight":         "output_norm.weight",
+	}
+
+	tMap := map[string]string{
+		"model.layers.(\\d+).input_layernorm.weight":                    "blk.$1.attn_norm.weight",
+		"model.layers.(\\d+).mlp.down_proj.weight":                      "blk.$1.ffn_down.weight",
+		"model.layers.(\\d+).mlp.gate_proj.weight":                      "blk.$1.ffn_gate.weight",
+		"model.layers.(\\d+).mlp.up_proj.weight":                        "blk.$1.ffn_up.weight",
+		"model.layers.(\\d+).post_attention_layernorm.weight":           "blk.$1.ffn_norm.weight",
+		"model.layers.(\\d+).self_attn.k_proj.weight":                   "blk.$1.attn_k.weight",
+		"model.layers.(\\d+).self_attn.o_proj.weight":                   "blk.$1.attn_output.weight",
+		"model.layers.(\\d+).self_attn.q_proj.weight":                   "blk.$1.attn_q.weight",
+		"model.layers.(\\d+).self_attn.v_proj.weight":                   "blk.$1.attn_v.weight",
+		"model.layers.(\\d+).block_sparse_moe.gate.weight":              "blk.$1.ffn_gate_inp.weight",
+		"model.layers.(\\d+).block_sparse_moe.experts.(\\d+).w1.weight": "blk.$1.ffn_gate.$2.weight",
+		"model.layers.(\\d+).block_sparse_moe.experts.(\\d+).w2.weight": "blk.$1.ffn_down.$2.weight",
+		"model.layers.(\\d+).block_sparse_moe.experts.(\\d+).w3.weight": "blk.$1.ffn_up.$2.weight",
+	}
+
+	v, ok := directMap[n]
+	if ok {
+		return v, nil
+	}
+
+	// quick hack to rename the layers to gguf format
+	for k, v := range tMap {
+		re := regexp.MustCompile(k)
+		newName := re.ReplaceAllString(n, v)
+		if newName != n {
+			return newName, nil
+		}
+	}
+
+	return "", fmt.Errorf("couldn't find a layer name for '%s'", n)
+}
+
+func (r safetensorWriterTo) WriteTo(w io.Writer) (n int64, err error) {
+	f, err := os.Open(r.filename)
+	if err != nil {
+		return 0, err
+	}
+	defer f.Close()
+
+	if _, err = f.Seek(r.offset, io.SeekStart); err != nil {
+		return 0, err
+	}
+
+	var f32s []float32
+	switch r.dtype {
+	case "F32":
+		f32s = make([]float32, r.size/4)
+		if err = binary.Read(f, r.bo, f32s); err != nil {
+			return 0, err
+		}
+	case "F16":
+		u16s := make([]uint16, r.size/2)
+		if err = binary.Read(f, r.bo, u16s); err != nil {
+			return 0, err
+		}
+
+		for _, b := range u16s {
+			f32s = append(f32s, float16.Frombits(b).Float32())
+		}
+
+	case "BF16":
+		u8s := make([]uint8, r.size)
+		if err = binary.Read(f, r.bo, u8s); err != nil {
+			return 0, err
+		}
+
+		f32s = bfloat16.DecodeFloat32(u8s)
+	default:
+		return 0, fmt.Errorf("unknown data type: %s", r.dtype)
+	}
+
+	if r.repacker != nil {
+		f32s, err = r.repacker(r.t.Name, f32s, r.t.Shape)
+		if err != nil {
+			return 0, err
+		}
+	}
+
+	switch r.t.Kind {
+	case 0:
+		return 0, binary.Write(w, r.bo, f32s)
+	case 1:
+		f16s := make([]uint16, len(f32s))
+		for i := range f32s {
+			f16s[i] = float16.Fromfloat32(f32s[i]).Bits()
+		}
+
+		return 0, binary.Write(w, r.bo, f16s)
+	default:
+		return 0, fmt.Errorf("unknown storage type: %d", r.t.Kind)
+	}
+}
+
+func (m *SafetensorFormat) GetModelArch(name, dirPath string, params *Params) (ModelArch, error) {
+	switch len(params.Architectures) {
+	case 0:
+		return nil, fmt.Errorf("No architecture specified to convert")
+	case 1:
+		switch params.Architectures[0] {
+		case "LlamaForCausalLM":
+			return &LlamaModel{
+				ModelData{
+					Name:   name,
+					Path:   dirPath,
+					Params: params,
+					Format: m,
+				},
+			}, nil
+		case "MistralForCausalLM":
+			return &MistralModel{
+				ModelData{
+					Name:   name,
+					Path:   dirPath,
+					Params: params,
+					Format: m,
+				},
+			}, nil
+		case "MixtralForCausalLM":
+			return &MixtralModel{
+				ModelData{
+					Name:   name,
+					Path:   dirPath,
+					Params: params,
+					Format: m,
+				},
+			}, nil
+		case "GemmaForCausalLM":
+			return &GemmaModel{
+				ModelData{
+					Name:   name,
+					Path:   dirPath,
+					Params: params,
+					Format: m,
+				},
+			}, nil
+		default:
+			return nil, fmt.Errorf("Models based on '%s' are not yet supported", params.Architectures[0])
+		}
+	}
+
+	return nil, fmt.Errorf("Unknown error")
+}

convert/sentencepiece_model.proto

@@ -0,0 +1,333 @@
+// Copyright 2016 Google Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.!
+
+syntax = "proto2";
+
+// TODO(taku): Needs to use LITE RUNTIME in OSS release.
+option optimize_for = LITE_RUNTIME;
+option go_package = "./sentencepiece";
+
+package sentencepiece;
+
+// TrainerSpec encodes a various parameters for SentencePiece training.
+// Next id: 55
+message TrainerSpec {
+  ///////////////////////////////////////////////////////////////////
+  // General parameters
+  //
+  // Input corpus files.
+  //  Trainer accepts the following two formats:
+  //  A) Monolingual: plain text, one sentence per line.
+  //  B) Bilingual:   TSV, source sentence <tab> target sentence
+  //  When bilingual data is passed, shared vocabulary model is built.
+  //  Note that the input file must be raw corpus, not a preprocessed corpus.
+  //  Trainer only loads the first `input_sentence_size` sentences specified
+  //  with this parameter.
+  repeated string input = 1;
+
+  // Input corpus format:
+  // "text": one-sentence-per-line text format (default)
+  // "tsv":  sentence <tab> freq
+  optional string input_format = 7;
+
+  // Output model file prefix.
+  // <model_prefix>.model and <model_prefix>.vocab are generated.
+  optional string model_prefix = 2;
+
+  // Model type. only have UNIGRAM now.
+  enum ModelType {
+    UNIGRAM = 1;  // Unigram language model with dynamic algorithm
+    BPE = 2;      // Byte Pair Encoding
+    WORD = 3;     // Delimitered by whitespace.
+    CHAR = 4;     // tokenizes into character sequence
+  }
+  optional ModelType model_type = 3 [default = UNIGRAM];
+
+  // Vocabulary size. 8k is the default size.
+  optional int32 vocab_size = 4 [default = 8000];
+
+  // List of the languages this model can accept.
+  // Since the model is language-agnostic, this field is used as a reference.
+  repeated string accept_language = 5;
+
+  // Size of self-test samples, which are encoded in the model file.
+  optional int32 self_test_sample_size = 6 [default = 0];
+
+  // Whether to use DP version of sentencepiece. Use it with TSV input format
+  // (requires precomputed word tab counts to work).
+  optional bool enable_differential_privacy = 50 [default = false];
+  // Set these parameters if you need DP version of sentencepiece.
+  // std of noise to add.
+  optional float differential_privacy_noise_level = 51 [default = 0.0];
+  // Clipping threshold to apply after adding noise. All the words with
+  // frequency less than this value are dropped.
+  optional uint64 differential_privacy_clipping_threshold = 52 [default = 0];
+
+  ///////////////////////////////////////////////////////////////////
+  // Training parameters.
+  //
+  // Uses characters which cover the corpus with the ratio of `chars_coverage`.
+  // This parameter determines the set of basic Alphabet of sentence piece.
+  // 1.0 - `chars_coverage` characters are treated as UNK.
+  // See also required_chars field.
+  optional float character_coverage = 10 [default = 0.9995];
+
+  // Maximum size of sentences the trainer loads from `input` parameter.
+  // Trainer simply loads the `input` files in sequence.
+  // It is better to shuffle the input corpus randomly.
+  optional uint64 input_sentence_size = 11 [default = 0];
+  optional bool shuffle_input_sentence = 19 [default = true];
+
+  // Maximum size of sentences to make seed sentence pieces.
+  // Extended suffix array is constructed to extract frequent
+  // sub-strings from the corpus. This uses 20N working space,
+  // where N is the size of corpus.
+  optional int32 mining_sentence_size = 12 [deprecated = true];
+
+  // Maximum size of sentences to train sentence pieces.
+  optional int32 training_sentence_size = 13 [deprecated = true];
+
+  // The size of seed sentencepieces.
+  // `seed_sentencepiece_size` must be larger than `vocab_size`.
+  optional int32 seed_sentencepiece_size = 14 [default = 1000000];
+
+  // In every EM sub-iterations, keeps top
+  // `shrinking_factor` * `current sentencepieces size` with respect to
+  // the loss of the sentence piece. This value should be smaller than 1.0.
+  optional float shrinking_factor = 15 [default = 0.75];
+
+  // The maximum sentence length in byte. The sentences with the length
+  // larger than `max_sentence_length` is simply ignored.
+  // Longer input tends to bring the following risks:
+  //  * Overflow during EM training (unigram language model only)
+  //  * Performance drop because of O(n log n) cost in BPE.
+  optional int32 max_sentence_length = 18 [default = 4192];
+
+  // Number of threads in the training.
+  optional int32 num_threads = 16 [default = 16];
+
+  // Number of EM sub iterations.
+  optional int32 num_sub_iterations = 17 [default = 2];
+
+  ///////////////////////////////////////////////////////////////////
+  // SentencePiece parameters which control the shapes of sentence piece.
+  //
+  // Maximum length of sentencepiece.
+  optional int32 max_sentencepiece_length = 20 [default = 16];
+
+  // Uses Unicode script to split sentence pieces.
+  // When `split_by_unicode_script` is true, we do not allow sentence piece to
+  // include multiple Unicode scripts, e.g. "F1" is not a valid piece.
+  // Exception: CJ characters (Hiragana/Katakana/Han) are all handled
+  // as one script type, since Japanese word can consist of multiple scripts.
+  // This exception is always applied regardless of the accept-language
+  // parameter.
+  optional bool split_by_unicode_script = 21 [default = true];
+
+  // When `split_by_number` is true, put a boundary between number and
+  // non-number transition. If we want to treat "F1" is one token, set this flag
+  // to be false.
+  optional bool split_by_number = 23 [default = true];
+
+  // Use a white space to split sentence pieces.
+  // When `split_by_whitespace` is false, we may have the piece containing
+  // a white space in the middle. e.g., "in_the".
+  optional bool split_by_whitespace = 22 [default = true];
+
+  // Adds whitespace symbol (_) as a suffix instead of prefix. e.g., _hello =>
+  // hello_. When `treat_whitespace_as_suffix` is true,
+  // NormalizerSpec::add_dummy_prefix will add the dummy whitespace to the end
+  // of sentence.
+  optional bool treat_whitespace_as_suffix = 24 [default = false];
+
+  // Allows pieces that only contain whitespaces instead of appearing only as
+  // prefix or suffix of other pieces.
+  optional bool allow_whitespace_only_pieces = 26 [default = false];
+
+  // Split all digits (0-9) into separate pieces.
+  optional bool split_digits = 25 [default = false];
+
+  // Defines the pre-tokenization delimiter.
+  // When specified, no pieces crossing this delimiter is not included
+  // in the vocab. Then the delimiter string is virtually ignored
+  // during the training. This field can allows constraints on the vocabulary
+  // selection. Note that this field is available on unigram mode.
+  optional string pretokenization_delimiter = 53 [ default = ""];
+
+  ///////////////////////////////////////////////////////////////////
+  // Vocabulary management
+  //
+  // Defines control symbols used as an indicator to
+  // change the behavior of the decoder. <s> and </s> are pre-defined.
+  // We can use this field to encode various meta information,
+  // including language indicator in multilingual model.
+  // These symbols are not visible to users, but visible to
+  // the decoder. Note that when the input sentence contains control symbols,
+  // they are not treated as one token, but segmented into normal pieces.
+  // Control symbols must be inserted independently from the segmentation.
+  repeated string control_symbols = 30;
+
+  // Defines user defined symbols.
+  // These symbols are added with extremely high score
+  // so they are always treated as one unique symbol in any context.
+  // Typical usage of user_defined_symbols is placeholder for named entities.
+  repeated string user_defined_symbols = 31;
+
+  // Defines required characters. Each UTF8 character in this string is included
+  // in the character set regardless of character_coverage value. Unlike
+  // user_defined_symbols, these characters have scores based on the frequency
+  // on input sentences, and the model can form subwords using characters
+  // in this field.
+  optional string required_chars = 36;
+
+  // Decomposes unknown pieces into UTF-8 bytes.
+  optional bool byte_fallback = 35 [default = false];
+
+  // When creating the vocabulary file, defines whether or not to additionally
+  // output the score for each piece.
+  optional bool vocabulary_output_piece_score = 32 [default = true];
+
+  // `vocab_size` is treated as hard limit. Crash if
+  // the model can not produce the vocab of size `vocab_size`,
+  // When `hard_vocab_limit` is false, vocab_size is treated
+  // as soft limit. Note that when model_type=char,
+  // always assumes hard_vocab_limit = false.
+  optional bool hard_vocab_limit = 33 [default = true];
+
+  // use all symbols for vocab extraction. This flag is valid
+  // if model type is either CHAR or WORD
+  optional bool use_all_vocab = 34 [default = false];
+
+  ///////////////////////////////////////////////////////////////////
+  // Reserved special meta tokens.
+  // * -1 is not used.
+  // * unk_id must not be -1.
+  // Id must starts with 0 and be contigous.
+  optional int32 unk_id = 40 [default = 0];   // <unk>
+  optional int32 bos_id = 41 [default = 1];   // <s>
+  optional int32 eos_id = 42 [default = 2];   // </s>
+  optional int32 pad_id = 43 [default = -1];  // <pad> (padding)
+  optional string unk_piece = 45 [default = "<unk>"];
+  optional string bos_piece = 46 [default = "<s>"];
+  optional string eos_piece = 47 [default = "</s>"];
+  optional string pad_piece = 48 [default = "<pad>"];
+
+  // Encodes <unk> into U+2047 (DOUBLE QUESTION MARK),
+  // since this character can be useful both for user and
+  // developer. We can easily figure out that <unk> is emitted.
+  optional string unk_surface = 44 [default = " \xE2\x81\x87 "];
+
+  // Increase bit depth to allow unigram model training on large
+  // (>10M sentences) corpora. A Side-effect of enabling this flag
+  // is increased memory usage.
+  optional bool train_extremely_large_corpus = 49 [default = false];
+
+ // Path to a seed sentencepieces file, with one tab-separated
+  // seed sentencepiece <tab> frequency per line.
+  optional string seed_sentencepieces_file = 54 [default = ""];
+
+  // Customized extensions: the range of field numbers
+  // are open to third-party extensions.
+  extensions 200 to max;
+}
+
+// NormalizerSpec encodes a various parameters for string normalizaiton
+message NormalizerSpec {
+  // name of normalization rule.
+  optional string name = 1;
+
+  // Pre-compiled normalization rule created by
+  // Builder::GetPrecompiledCharsMap() or Builder::CompileCharsMap() method.
+  // Usually this field is set by Builder::GetNormalizerSpec() method.
+  optional bytes precompiled_charsmap = 2;
+
+  // Adds dummy whitespace at the beginning of text in order to
+  // treat "world" in "world" and "hello world" in the same way.
+  optional bool add_dummy_prefix = 3 [default = true];
+
+  // Removes leading, trailing, and duplicate internal whitespace.
+  optional bool remove_extra_whitespaces = 4 [default = true];
+
+  // Replaces whitespace with meta symbol.
+  // This field must be true to train sentence piece model.
+  optional bool escape_whitespaces = 5 [default = true];
+
+  // Custom normalization rule file in TSV format.
+  // https://github.com/google/sentencepiece/blob/master/doc/normalization.md
+  // This field is only used in SentencePieceTrainer::Train() method, which
+  // compiles the rule into the binary rule stored in `precompiled_charsmap`.
+  optional string normalization_rule_tsv = 6;
+
+  // Customized extensions: the range of field numbers
+  // are open to third-party extensions.
+  extensions 200 to max;
+}
+
+// Proto to store samples for self-testing.
+message SelfTestData {
+  message Sample {
+    optional string input = 1;
+    optional string expected = 2;
+  }
+  repeated Sample samples = 1;
+
+  // Customized extensions: the range of field numbers
+  // are open to third-party extensions.
+  extensions 200 to max;
+}
+
+// ModelProto stores model parameters.
+// SentencePieceProcessor is supposed to be self-contained.
+// All settings/parameters which may change the behavior must be encoded
+// in ModelProto.
+message ModelProto {
+  message SentencePiece {
+    enum Type {
+      NORMAL = 1;        // normal symbol
+      UNKNOWN = 2;       // unknown symbol. only <unk> for now.
+      CONTROL = 3;       // control symbols. </s>, <s>, <2ja> etc.
+      USER_DEFINED = 4;  // user defined symbols.
+                         // Typical usage of USER_DEFINED symbol
+                         // is placeholder.
+      BYTE = 6;          // byte symbols. Used when `byte_fallback` is true.
+      UNUSED = 5;        // this piece is not used.
+    }
+    optional string piece = 1;  // piece must not be empty.
+    optional float score = 2;
+    optional Type type = 3 [default = NORMAL];
+
+    // Customized extensions: the range of field numbers
+    // are open to third-party extensions.
+    extensions 200 to max;
+  }
+
+  // Sentence pieces with scores.
+  repeated SentencePiece pieces = 1;
+
+  // Spec used to generate this model file.
+  optional TrainerSpec trainer_spec = 2;
+
+  // Spec for text normalization.
+  optional NormalizerSpec normalizer_spec = 3;
+
+  // Stores sample input and its expected segmentation to verify the model.
+  optional SelfTestData self_test_data = 4;
+
+  // Spec for text de-normalization.
+  optional NormalizerSpec denormalizer_spec = 5;
+
+  // Customized extensions: the range of field numbers
+  // are open to third-party extensions.
+  extensions 200 to max;
+}

convert/tokenizer.go

@@ -0,0 +1,106 @@
+package convert
+
+import (
+	"cmp"
+	"crypto/sha256"
+	"encoding/json"
+	"fmt"
+	"log/slog"
+	"os"
+	"slices"
+
+	"golang.org/x/exp/maps"
+)
+
+type Tokenizer struct {
+	Version     string         `json:"version"`
+	AddedTokens []Token        `json:"added_tokens"`
+	Model       TokenizerModel `json:"model"`
+
+	PreTokenizer struct {
+		PreTokenizers []struct {
+			Type    string `json:"type"`
+			Pattern struct {
+				Regex string `json:"Regex"`
+			} `json:"pattern"`
+		} `json:"pretokenizers"`
+	} `json:"pre_tokenizer"`
+}
+
+type TokenizerModel struct {
+	Type   string         `json:"type"`
+	Vocab  map[string]int `json:"vocab"`
+	Merges []string       `json:"merges"`
+	Tokens []Token
+}
+
+type Token struct {
+	ID          int    `json:"id"`
+	Content     string `json:"content"`
+	Special     bool   `json:"special"`
+	UserDefined bool
+}
+
+func (t *Token) Type() int32 {
+	switch {
+	case t.Special:
+		return tokenTypeControl
+	case t.UserDefined:
+		return tokenTypeUserDefined
+	default:
+		return tokenTypeNormal
+	}
+}
+
+func (t *Tokenizer) maxID() int {
+	return max(
+		slices.Max(maps.Values(t.Model.Vocab)),
+		slices.MaxFunc(t.AddedTokens, func(a, b Token) int {
+			return cmp.Compare(a.ID, b.ID)
+		}).ID,
+	)
+}
+
+func parseTokens(dirpath string) (pre string, tokens []Token, merges []string, err error) {
+	f, err := os.Open(dirpath)
+	if err != nil {
+		panic(err)
+	}
+	defer f.Close()
+
+	var t Tokenizer
+	if err := json.NewDecoder(f).Decode(&t); err != nil {
+		return "", nil, nil, err
+	}
+
+	tokens = make([]Token, t.maxID()+1)
+	for k, v := range t.Model.Vocab {
+		tokens[v] = Token{ID: v, Content: k, Special: false, UserDefined: false}
+	}
+
+	for _, v := range t.AddedTokens {
+		v.UserDefined = true
+		tokens[v.ID] = v
+	}
+
+	sha256sum := sha256.New()
+	for _, pt := range t.PreTokenizer.PreTokenizers {
+		if pt.Type == "Split" && pt.Pattern.Regex != "" {
+			sha256sum.Write([]byte(pt.Pattern.Regex))
+		}
+	}
+
+	switch digest := fmt.Sprintf("%x", sha256sum.Sum(nil)); digest {
+	case "d98f9631be1e9607a9848c26c1f9eac1aa9fc21ac6ba82a2fc0741af9780a48f":
+		pre = "llama-bpe"
+	case "03df5c5863ad70781dcfdef491ead25140f895fe8010964be0daefe27be32b02":
+		pre = "deepseek-llm"
+	case "21cde974d587f0d54dc8d56b183cc1e6239600172035c68fbd6d4b9f8da0576e":
+		pre = "deepseek-coder"
+	default:
+		slog.Warn("unknown pretokenizer, using default", "digest", digest)
+		pre = "default"
+	}
+
+	return pre, tokens, t.Model.Merges, nil
+}

convert/torch.go

@@ -0,0 +1,287 @@
+package convert
+
+import (
+	"encoding/binary"
+	"encoding/json"
+	"fmt"
+	"io"
+	"log/slog"
+	"os"
+	"path/filepath"
+	"regexp"
+	"strings"
+
+	"github.com/nlpodyssey/gopickle/pytorch"
+	"github.com/nlpodyssey/gopickle/types"
+	"github.com/x448/float16"
+
+	"github.com/ollama/ollama/llm"
+)
+
+type torchWriterTo struct {
+	t *llm.Tensor
+
+	params *Params
+	bo     ByteOrder
+
+	storage  pytorch.StorageInterface
+	repacker func(string, []float32, []uint64) ([]float32, error)
+}
+
+type TorchFormat struct{}
+
+func (tf *TorchFormat) GetTensors(dirpath string, params *Params) ([]llm.Tensor, error) {
+	slog.Debug("getting torch tensors")
+
+	var files []string
+	if pt, _ := filepath.Glob(filepath.Join(dirpath, "consolidated*.pth")); len(pt) > 0 {
+		files = append(files, pt...)
+	} else if pt, _ := filepath.Glob(filepath.Join(dirpath, "pytorch_model*.pth")); len(pt) > 0 {
+		files = append(files, pt...)
+	}
+
+	var offset uint64
+	var tensors []llm.Tensor
+	for _, fn := range files {
+		m, err := pytorch.Load(fn)
+		if err != nil {
+			slog.Error(fmt.Sprintf("error unpickling: %q", err))
+			return []llm.Tensor{}, err
+		}
+
+		for _, k := range m.(*types.Dict).Keys() {
+			if strings.HasSuffix(k.(string), "self_attn.rotary_emb.inv_freq") {
+				continue
+			}
+
+			t, _ := m.(*types.Dict).Get(k)
+			tshape := t.(*pytorch.Tensor).Size
+
+			var size uint64
+			var kind uint32
+			switch len(tshape) {
+			case 0:
+				continue
+			case 1:
+				// convert to float32
+				kind = 0
+				size = uint64(tshape[0] * 4)
+			case 2:
+				// convert to float16
+				kind = 1
+				size = uint64(tshape[0] * tshape[1] * 2)
+			}
+
+			ggufName, err := tf.GetLayerName(k.(string))
+			if err != nil {
+				slog.Error(err.Error())
+				return nil, err
+			}
+			slog.Debug(fmt.Sprintf("'%35s': '%30s' %10d [%#v]", k.(string), ggufName, size, tshape))
+
+			shape := []uint64{0, 0, 0, 0}
+			for i := range tshape {
+				shape[i] = uint64(tshape[i])
+			}
+
+			tensor := llm.Tensor{
+				Name:   ggufName,
+				Kind:   kind,
+				Offset: offset, // calculate the offset
+				Shape:  shape,
+			}
+
+			tensor.WriterTo = torchWriterTo{
+				t:       &tensor,
+				params:  params,
+				bo:      params.ByteOrder,
+				storage: t.(*pytorch.Tensor).Source,
+			}
+
+			tensors = append(tensors, tensor)
+			offset += size
+		}
+	}
+
+	return tensors, nil
+}
+
+func getAltParams(dirpath string) (*Params, error) {
+	f, err := os.Open(filepath.Join(dirpath, "params.json"))
+	if err != nil {
+		slog.Error("no params.json")
+		return nil, err
+	}
+	defer f.Close()
+
+	type TorchParams struct {
+		HiddenSize     int     `json:"dim"`
+		AttentionHeads int     `json:"n_heads"`
+		KeyValHeads    int     `json:"n_kv_heads"`
+		HiddenLayers   int     `json:"n_layers"`
+		RopeTheta      float64 `json:"rope_theta"`
+		NormEPS        float64 `json:"norm_eps"`
+	}
+
+	var tparams TorchParams
+
+	d := json.NewDecoder(f)
+	err = d.Decode(&tparams)
+	if err != nil {
+		return nil, err
+	}
+
+	params := &Params{
+		Architectures:  []string{"LlamaForCausalLM"},
+		HiddenSize:     tparams.HiddenSize,
+		AttentionHeads: tparams.AttentionHeads,
+		KeyValHeads:    tparams.KeyValHeads,
+		HiddenLayers:   tparams.HiddenLayers,
+		NormEPS:        tparams.NormEPS,
+	}
+
+	switch {
+	case tparams.RopeTheta == 1000000:
+		// Codellama
+		params.ContextSize = 16384
+	case tparams.NormEPS == 1e-06:
+		// llama2
+		slog.Debug("Found llama2 - setting context size to 4096")
+		params.ContextSize = 4096
+	default:
+		params.ContextSize = 2048
+	}
+
+	params.ByteOrder = binary.LittleEndian
+	return params, nil
+}
+
+func (m *TorchFormat) GetParams(dirpath string) (*Params, error) {
+	f, err := os.Open(filepath.Join(dirpath, "config.json"))
+	if err != nil {
+		if os.IsNotExist(err) {
+			// try params.json instead
+			return getAltParams(dirpath)
+		} else {
+			return nil, err
+		}
+	}
+
+	var params Params
+	d := json.NewDecoder(f)
+	err = d.Decode(&params)
+	if err != nil {
+		return nil, err
+	}
+
+	params.ByteOrder = binary.LittleEndian
+	return &params, nil
+}
+
+func (m *TorchFormat) GetLayerName(n string) (string, error) {
+	directMap := map[string]string{
+		"tok_embeddings.weight":     "token_embd.weight",
+		"output.weight":             "output.weight",
+		"norm.weight":               "output_norm.weight",
+		"rope.freqs":                "rope_freqs.weight",
+		"model.embed_tokens.weight": "token_embd.weight",
+		"lm_head.weight":            "output.weight",
+		"model.norm.weight":         "output_norm.weight",
+	}
+
+	lMap := map[string]string{
+		"layers.(\\d+).attention_norm.weight":                 "blk.$1.attn_norm.weight",
+		"layers.(\\d+).attention_output_norm.weight":          "blk.$1.attn_norm.weight",
+		"layers.(\\d+).feed_forward.w2.weight":                "blk.$1.ffn_down.weight",
+		"layers.(\\d+).feed_forward.w1.weight":                "blk.$1.ffn_gate.weight",
+		"layers.(\\d+).feed_forward.w3.weight":                "blk.$1.ffn_up.weight",
+		"layers.(\\d+).ffn_norm.weight":                       "blk.$1.ffn_norm.weight",
+		"layers.(\\d+).attention.wk.weight":                   "blk.$1.attn_k.weight",
+		"layers.(\\d+).attention.wo.weight":                   "blk.$1.attn_output.weight",
+		"layers.(\\d+).attention.wq.weight":                   "blk.$1.attn_q.weight",
+		"layers.(\\d+).attention.wv.weight":                   "blk.$1.attn_v.weight",
+		"model.layers.(\\d+).input_layernorm.weight":          "blk.$1.attn_norm.weight",
+		"model.layers.(\\d+).mlp.down_proj.weight":            "blk.$1.ffn_down.weight",
+		"model.layers.(\\d+).mlp.gate_proj.weight":            "blk.$1.ffn_gate.weight",
+		"model.layers.(\\d+).mlp.up_proj.weight":              "blk.$1.ffn_up.weight",
+		"model.layers.(\\d+).post_attention_layernorm.weight": "blk.$1.ffn_norm.weight",
+		"model.layers.(\\d+).self_attn.k_proj.weight":         "blk.$1.attn_k.weight",
+		"model.layers.(\\d+).self_attn.o_proj.weight":         "blk.$1.attn_output.weight",
+		"model.layers.(\\d+).self_attn.q_proj.weight":         "blk.$1.attn_q.weight",
+		"model.layers.(\\d+).self_attn.v_proj.weight":         "blk.$1.attn_v.weight",
+	}
+
+	v, ok := directMap[n]
+	if ok {
+		return v, nil
+	}
+
+	// quick hack to rename the layers to gguf format
+	for k, v := range lMap {
+		re := regexp.MustCompile(k)
+		newName := re.ReplaceAllString(n, v)
+		if newName != n {
+			return newName, nil
+		}
+	}
+
+	return "", fmt.Errorf("couldn't find a layer name for '%s'", n)
+}
+
+func (r torchWriterTo) WriteTo(w io.Writer) (n int64, err error) {
+	var f32s []float32
+	switch s := r.storage.(type) {
+	case *pytorch.FloatStorage:
+		f32s = s.Data
+	case *pytorch.HalfStorage:
+		f32s = s.Data
+	case *pytorch.BFloat16Storage:
+		f32s = s.Data
+	default:
+		return 0, fmt.Errorf("unknown data type: %T", s)
+	}
+
+	if r.repacker != nil {
+		f32s, err = r.repacker(r.t.Name, f32s, r.t.Shape)
+		if err != nil {
+			return 0, err
+		}
+	}
+
+	switch r.t.Kind {
+	case 0:
+		return 0, binary.Write(w, r.bo, f32s)
+	case 1:
+		f16s := make([]uint16, len(f32s))
+		for i := range f32s {
+			f16s[i] = float16.Fromfloat32(f32s[i]).Bits()
+		}
+
+		return 0, binary.Write(w, r.bo, f16s)
+	default:
+		return 0, fmt.Errorf("unknown storage type: %d", r.t.Kind)
+	}
+}
+
+func (m *TorchFormat) GetModelArch(name, dirPath string, params *Params) (ModelArch, error) {
+	switch len(params.Architectures) {
+	case 0:
+		return nil, fmt.Errorf("No architecture specified to convert")
+	case 1:
+		switch params.Architectures[0] {
+		case "LlamaForCausalLM":
+			return &LlamaModel{
+				ModelData{
+					Name:   name,
+					Path:   dirPath,
+					Params: params,
+					Format: m,
+				},
+			}, nil
+		default:
+			return nil, fmt.Errorf("Models based on '%s' are not yet supported", params.Architectures[0])
+		}
+	}
+
+	return nil, fmt.Errorf("Unknown error")
+}

docs/README.md

@@ -1,25 +1,21 @@
 # Documentation
 
-To get started, see the project's **[quickstart](../README.md#quickstart)**.
+### Getting Started
+* [Quickstart](../README.md#quickstart)
+* [Examples](../examples)
+* [Importing models](./import.md)
+* [Linux Documentation](./linux.md)
+* [Windows Documentation](./windows.md)
+* [Docker Documentation](./docker.md)
 
-Ollama is a tool for running AI models on your hardware. Many users will choose to use the Command Line Interface (CLI) to work with Ollama. Learn more about all the commands in the CLI in the **[Main Readme](../README.md)**.
+### Reference
 
-Use the RESTful API using any language, including Python, JavaScript, Typescript, Go, Rust, and many more. Learn more about using the API in the **[API Documentation](./api.md)**.
+* [API Reference](./api.md)
+* [Modelfile Reference](./modelfile.md)
+* [OpenAI Compatibility](./openai.md)
 
-Create new models or modify models already in the library using the Modelfile. Learn more about the Modelfile syntax in the **[Modelfile Documentation](./modelfile.md)**.
+### Resources
 
-Import models using source model weights found on Hugging Face and similar sites by referring to the **[Import Documentation](./import.md)**.
-
-Installing on Linux in most cases is easy using the script on [ollama.com/download](ollama.com/download). To get more detail about the install, including CUDA drivers, see the **[Linux Documentation](./linux.md)**.
-
-Many of our users like the flexibility of using our official Docker Image. Learn more about using Docker with Ollama using the **[Docker Documentation](https://hub.docker.com/r/ollama/ollama)**.
-
-It is easy to install on Linux and Mac, but many users will choose to build Ollama on their own. To do this, refer to the **[Development Documentation](./development.md)**.
-
-If encountering a problem with Ollama, the best place to start is the logs. Find more information about them here in the **[Troubleshooting Guide](./troubleshooting.md)**.
-
-Finally for all the questions that don't fit anywhere else, there is the **[FAQ](./faq.md)**
-
-[Tutorials](./tutorials.md) apply the documentation to tasks.
-
-For working code examples of using Ollama, see [Examples](../examples).
+* [Troubleshooting Guide](./troubleshooting.md)
+* [FAQ](./faq.md)
+* [Development guide](./development.md)

docs/api.md

@@ -12,12 +12,13 @@
 - [Pull a Model](#pull-a-model)
 - [Push a Model](#push-a-model)
 - [Generate Embeddings](#generate-embeddings)
+- [List Running Models](#list-running-models)
 
 ## Conventions
 
 ### Model names
 
-Model names follow a `model:tag` format, where `model` can have an optional namespace such as `example/model`. Some examples are `orca-mini:3b-q4_1` and `llama2:70b`. The tag is optional and, if not provided, will default to `latest`. The tag is used to identify a specific version.
+Model names follow a `model:tag` format, where `model` can have an optional namespace such as `example/model`. Some examples are `orca-mini:3b-q4_1` and `llama3:70b`. The tag is optional and, if not provided, will default to `latest`. The tag is used to identify a specific version.
 
 ### Durations
 
@@ -25,7 +26,7 @@ All durations are returned in nanoseconds.
 
 ### Streaming responses
 
-Certain endpoints stream responses as JSON objects and can optional return non-streamed responses.
+Certain endpoints stream responses as JSON objects. Streaming can be disabled by providing `{"stream": false}` for these endpoints.
 
 ## Generate a completion
 
@@ -54,7 +55,7 @@ Advanced parameters (optional):
 
 #### JSON mode
 
-Enable JSON mode by setting the `format` parameter to `json`. This will structure the response as a valid JSON object. See the JSON mode [example](#generate-request-json-mode) below.
+Enable JSON mode by setting the `format` parameter to `json`. This will structure the response as a valid JSON object. See the JSON mode [example](#request-json-mode) below.
 
 > Note: it's important to instruct the model to use JSON in the `prompt`. Otherwise, the model may generate large amounts whitespace.
 
@@ -66,7 +67,7 @@ Enable JSON mode by setting the `format` parameter to `json`. This will structur
 
 ```shell
 curl http://localhost:11434/api/generate -d '{
-  "model": "llama2",
+  "model": "llama3",
   "prompt": "Why is the sky blue?"
 }'
 ```
@@ -77,7 +78,7 @@ A stream of JSON objects is returned:
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-08-04T08:52:19.385406455-07:00",
   "response": "The",
   "done": false
@@ -90,16 +91,16 @@ The final response in the stream also includes additional data about the generat
 - `load_duration`: time spent in nanoseconds loading the model
 - `prompt_eval_count`: number of tokens in the prompt
 - `prompt_eval_duration`: time spent in nanoseconds evaluating the prompt
-- `eval_count`: number of tokens the response
+- `eval_count`: number of tokens in the response
 - `eval_duration`: time in nanoseconds spent generating the response
 - `context`: an encoding of the conversation used in this response, this can be sent in the next request to keep a conversational memory
 - `response`: empty if the response was streamed, if not streamed, this will contain the full response
 
-To calculate how fast the response is generated in tokens per second (token/s), divide `eval_count` / `eval_duration`.
+To calculate how fast the response is generated in tokens per second (token/s), divide `eval_count` / `eval_duration` * `10^9`.
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-08-04T19:22:45.499127Z",
   "response": "",
   "done": true,
@@ -121,7 +122,7 @@ A response can be received in one reply when streaming is off.
 
 ```shell
 curl http://localhost:11434/api/generate -d '{
-  "model": "llama2",
+  "model": "llama3",
   "prompt": "Why is the sky blue?",
   "stream": false
 }'
@@ -133,7 +134,7 @@ If `stream` is set to `false`, the response will be a single JSON object:
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-08-04T19:22:45.499127Z",
   "response": "The sky is blue because it is the color of the sky.",
   "done": true,
@@ -155,7 +156,7 @@ If `stream` is set to `false`, the response will be a single JSON object:
 
 ```shell
 curl http://localhost:11434/api/generate -d '{
-  "model": "llama2",
+  "model": "llama3",
   "prompt": "What color is the sky at different times of the day? Respond using JSON",
   "format": "json",
   "stream": false
@@ -166,7 +167,7 @@ curl http://localhost:11434/api/generate -d '{
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-11-09T21:07:55.186497Z",
   "response": "{\n\"morning\": {\n\"color\": \"blue\"\n},\n\"noon\": {\n\"color\": \"blue-gray\"\n},\n\"afternoon\": {\n\"color\": \"warm gray\"\n},\n\"evening\": {\n\"color\": \"orange\"\n}\n}\n",
   "done": true,
@@ -249,17 +250,16 @@ curl http://localhost:11434/api/generate -d '{
 
 #### Request (Reproducible outputs)
 
-For reproducible outputs, set `temperature` to 0 and `seed` to a number:
+For reproducible outputs, set `seed` to a number:
 
 ##### Request
 
 ```shell
 curl http://localhost:11434/api/generate -d '{
   "model": "mistral",
-  "prompt": "[INST] why is the sky blue? [/INST]",
+  "prompt": "Why is the sky blue?",
   "options": {
-    "seed": 101,
-    "temperature": 0
+    "seed": 123
   }
 }'
 ```
@@ -289,7 +289,7 @@ If you want to set custom options for the model at runtime rather than in the Mo
 
 ```shell
 curl http://localhost:11434/api/generate -d '{
-  "model": "llama2",
+  "model": "llama3",
   "prompt": "Why is the sky blue?",
   "stream": false,
   "options": {
@@ -313,7 +313,6 @@ curl http://localhost:11434/api/generate -d '{
     "numa": false,
     "num_ctx": 1024,
     "num_batch": 2,
-    "num_gqa": 1,
     "num_gpu": 1,
     "main_gpu": 0,
     "low_vram": false,
@@ -321,9 +320,6 @@ curl http://localhost:11434/api/generate -d '{
     "vocab_only": false,
     "use_mmap": true,
     "use_mlock": false,
-    "embedding_only": false,
-    "rope_frequency_base": 1.1,
-    "rope_frequency_scale": 0.8,
     "num_thread": 8
   }
 }'
@@ -333,7 +329,7 @@ curl http://localhost:11434/api/generate -d '{
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-08-04T19:22:45.499127Z",
   "response": "The sky is blue because it is the color of the sky.",
   "done": true,
@@ -355,7 +351,7 @@ If an empty prompt is provided, the model will be loaded into memory.
 
 ```shell
 curl http://localhost:11434/api/generate -d '{
-  "model": "llama2"
+  "model": "llama3"
 }'
 ```
 
@@ -365,7 +361,7 @@ A single JSON object is returned:
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-12-18T19:52:07.071755Z",
   "response": "",
   "done": true
@@ -395,7 +391,6 @@ Advanced parameters (optional):
 
 - `format`: the format to return a response in. Currently the only accepted value is `json`
 - `options`: additional model parameters listed in the documentation for the [Modelfile](./modelfile.md#valid-parameters-and-values) such as `temperature`
-- `template`: the prompt template to use (overrides what is defined in the `Modelfile`)
 - `stream`: if `false` the response will be returned as a single response object, rather than a stream of objects
 - `keep_alive`: controls how long the model will stay loaded into memory following the request (default: `5m`)
 
@@ -409,7 +404,7 @@ Send a chat message with a streaming response.
 
 ```shell
 curl http://localhost:11434/api/chat -d '{
-  "model": "llama2",
+  "model": "llama3",
   "messages": [
     {
       "role": "user",
@@ -425,7 +420,7 @@ A stream of JSON objects is returned:
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-08-04T08:52:19.385406455-07:00",
   "message": {
     "role": "assistant",
@@ -440,7 +435,7 @@ Final response:
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-08-04T19:22:45.499127Z",
   "done": true,
   "total_duration": 4883583458,
@@ -458,7 +453,7 @@ Final response:
 
 ```shell
 curl http://localhost:11434/api/chat -d '{
-  "model": "llama2",
+  "model": "llama3",
   "messages": [
     {
       "role": "user",
@@ -473,7 +468,7 @@ curl http://localhost:11434/api/chat -d '{
 
 ```json
 {
-  "model": "registry.ollama.ai/library/llama2:latest",
+  "model": "registry.ollama.ai/library/llama3:latest",
   "created_at": "2023-12-12T14:13:43.416799Z",
   "message": {
     "role": "assistant",
@@ -497,7 +492,7 @@ Send a chat message with a conversation history. You can use this same approach
 
 ```shell
 curl http://localhost:11434/api/chat -d '{
-  "model": "llama2",
+  "model": "llama3",
   "messages": [
     {
       "role": "user",
@@ -521,7 +516,7 @@ A stream of JSON objects is returned:
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-08-04T08:52:19.385406455-07:00",
   "message": {
     "role": "assistant",
@@ -535,7 +530,7 @@ Final response:
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-08-04T19:22:45.499127Z",
   "done": true,
   "total_duration": 8113331500,
@@ -593,7 +588,7 @@ curl http://localhost:11434/api/chat -d '{
 
 ```shell
 curl http://localhost:11434/api/chat -d '{
-  "model": "llama2",
+  "model": "llama3",
   "messages": [
     {
       "role": "user",
@@ -611,7 +606,7 @@ curl http://localhost:11434/api/chat -d '{
 
 ```json
 {
-  "model": "registry.ollama.ai/library/llama2:latest",
+  "model": "registry.ollama.ai/library/llama3:latest",
   "created_at": "2023-12-12T14:13:43.416799Z",
   "message": {
     "role": "assistant",
@@ -653,7 +648,7 @@ Create a new model from a `Modelfile`.
 ```shell
 curl http://localhost:11434/api/create -d '{
   "name": "mario",
-  "modelfile": "FROM llama2\nSYSTEM You are mario from Super Mario Bros."
+  "modelfile": "FROM llama3\nSYSTEM You are mario from Super Mario Bros."
 }'
 ```
 
@@ -760,7 +755,7 @@ A single JSON object will be returned.
       }
     },
     {
-      "name": "llama2:latest",
+      "name": "llama3:latest",
       "modified_at": "2023-12-07T09:32:18.757212583-08:00",
       "size": 3825819519,
       "digest": "fe938a131f40e6f6d40083c9f0f430a515233eb2edaa6d72eb85c50d64f2300e",
@@ -782,11 +777,12 @@ A single JSON object will be returned.
 POST /api/show
 ```
 
-Show information about a model including details, modelfile, template, parameters, license, and system prompt.
+Show information about a model including details, modelfile, template, parameters, license, system prompt.
 
 ### Parameters
 
 - `name`: name of the model to show
+- `verbose`: (optional) if set to `true`, returns full data for verbose response fields
 
 ### Examples
 
@@ -794,7 +790,7 @@ Show information about a model including details, modelfile, template, parameter
 
 ```shell
 curl http://localhost:11434/api/show -d '{
-  "name": "llama2"
+  "name": "llama3"
 }'
 ```
 
@@ -802,15 +798,41 @@ curl http://localhost:11434/api/show -d '{
 
 ```json
 {
-  "modelfile": "# Modelfile generated by \"ollama show\"\n# To build a new Modelfile based on this one, replace the FROM line with:\n# FROM llava:latest\n\nFROM /Users/matt/.ollama/models/blobs/sha256:200765e1283640ffbd013184bf496e261032fa75b99498a9613be4e94d63ad52\nTEMPLATE \"\"\"{{ .System }}\nUSER: {{ .Prompt }}\nASSSISTANT: \"\"\"\nPARAMETER num_ctx 4096\nPARAMETER stop \"\u003c/s\u003e\"\nPARAMETER stop \"USER:\"\nPARAMETER stop \"ASSSISTANT:\"",
-  "parameters": "num_ctx                        4096\nstop                           \u003c/s\u003e\nstop                           USER:\nstop                           ASSSISTANT:",
-  "template": "{{ .System }}\nUSER: {{ .Prompt }}\nASSSISTANT: ",
+  "modelfile": "# Modelfile generated by \"ollama show\"\n# To build a new Modelfile based on this one, replace the FROM line with:\n# FROM llava:latest\n\nFROM /Users/matt/.ollama/models/blobs/sha256:200765e1283640ffbd013184bf496e261032fa75b99498a9613be4e94d63ad52\nTEMPLATE \"\"\"{{ .System }}\nUSER: {{ .Prompt }}\nASSISTANT: \"\"\"\nPARAMETER num_ctx 4096\nPARAMETER stop \"\u003c/s\u003e\"\nPARAMETER stop \"USER:\"\nPARAMETER stop \"ASSISTANT:\"",
+  "parameters": "num_keep                       24\nstop                           \"<|start_header_id|>\"\nstop                           \"<|end_header_id|>\"\nstop                           \"<|eot_id|>\"",
+  "template": "{{ if .System }}<|start_header_id|>system<|end_header_id|>\n\n{{ .System }}<|eot_id|>{{ end }}{{ if .Prompt }}<|start_header_id|>user<|end_header_id|>\n\n{{ .Prompt }}<|eot_id|>{{ end }}<|start_header_id|>assistant<|end_header_id|>\n\n{{ .Response }}<|eot_id|>",
   "details": {
+    "parent_model": "",
     "format": "gguf",
     "family": "llama",
-    "families": ["llama", "clip"],
-    "parameter_size": "7B",
+    "families": [
+      "llama"
+    ],
+    "parameter_size": "8.0B",
     "quantization_level": "Q4_0"
+  },
+  "model_info": {
+    "general.architecture": "llama",
+    "general.file_type": 2,
+    "general.parameter_count": 8030261248,
+    "general.quantization_version": 2,
+    "llama.attention.head_count": 32,
+    "llama.attention.head_count_kv": 8,
+    "llama.attention.layer_norm_rms_epsilon": 0.00001,
+    "llama.block_count": 32,
+    "llama.context_length": 8192,
+    "llama.embedding_length": 4096,
+    "llama.feed_forward_length": 14336,
+    "llama.rope.dimension_count": 128,
+    "llama.rope.freq_base": 500000,
+    "llama.vocab_size": 128256,
+    "tokenizer.ggml.bos_token_id": 128000,
+    "tokenizer.ggml.eos_token_id": 128009,
+    "tokenizer.ggml.merges": [],            // populates if `verbose=true`
+    "tokenizer.ggml.model": "gpt2",
+    "tokenizer.ggml.pre": "llama-bpe",
+    "tokenizer.ggml.token_type": [],        // populates if `verbose=true`
+    "tokenizer.ggml.tokens": []             // populates if `verbose=true`
   }
 }
 ```
@@ -829,8 +851,8 @@ Copy a model. Creates a model with another name from an existing model.
 
 ```shell
 curl http://localhost:11434/api/copy -d '{
-  "source": "llama2",
-  "destination": "llama2-backup"
+  "source": "llama3",
+  "destination": "llama3-backup"
 }'
 ```
 
@@ -856,7 +878,7 @@ Delete a model and its data.
 
 ```shell
 curl -X DELETE http://localhost:11434/api/delete -d '{
-  "name": "llama2:13b"
+  "name": "llama3:13b"
 }'
 ```
 
@@ -884,7 +906,7 @@ Download a model from the ollama library. Cancelled pulls are resumed from where
 
 ```shell
 curl http://localhost:11434/api/pull -d '{
-  "name": "llama2"
+  "name": "llama3"
 }'
 ```
 
@@ -1025,7 +1047,7 @@ Advanced parameters:
 
 ```shell
 curl http://localhost:11434/api/embeddings -d '{
-  "model": "llama2",
+  "model": "all-minilm",
   "prompt": "Here is an article about llamas..."
 }'
 ```
@@ -1040,3 +1062,47 @@ curl http://localhost:11434/api/embeddings -d '{
   ]
 }
 ```
+
+## List Running Models
+```shell
+GET /api/ps
+```
+
+List models that are currently loaded into memory.
+
+#### Examples
+
+### Request
+
+```shell
+curl http://localhost:11434/api/ps
+```
+
+#### Response
+
+A single JSON object will be returned.
+
+```json
+{
+  "models": [
+    {
+      "name": "mistral:latest",
+      "model": "mistral:latest",
+      "size": 5137025024,
+      "digest": "2ae6f6dd7a3dd734790bbbf58b8909a606e0e7e97e94b7604e0aa7ae4490e6d8",
+      "details": {
+        "parent_model": "",
+        "format": "gguf",
+        "family": "llama",
+        "families": [
+          "llama"
+        ],
+        "parameter_size": "7.2B",
+        "quantization_level": "Q4_0"
+      },
+      "expires_at": "2024-06-04T14:38:31.83753-07:00",
+      "size_vram": 5137025024
+    }
+  ]
+}
+```

docs/development.md

@@ -3,9 +3,11 @@
 Install required tools:
 
 - cmake version 3.24 or higher
-- go version 1.21 or higher
+- go version 1.22 or higher
 - gcc version 11.4.0 or higher
 
+### MacOS
+
 ```bash
 brew install go cmake gcc
 ```
@@ -42,16 +44,16 @@ Now you can run `ollama`:
 
 #### Linux CUDA (NVIDIA)
 
-*Your operating system distribution may already have packages for NVIDIA CUDA. Distro packages are often preferable, but instructions are distro-specific. Please consult distro-specific docs for dependencies if available!*
+_Your operating system distribution may already have packages for NVIDIA CUDA. Distro packages are often preferable, but instructions are distro-specific. Please consult distro-specific docs for dependencies if available!_
 
 Install `cmake` and `golang` as well as [NVIDIA CUDA](https://developer.nvidia.com/cuda-downloads)
-development and runtime packages. 
+development and runtime packages.
 
 Typically the build scripts will auto-detect CUDA, however, if your Linux distro
 or installation approach uses unusual paths, you can specify the location by
 specifying an environment variable `CUDA_LIB_DIR` to the location of the shared
-libraries, and `CUDACXX` to the location of the nvcc compiler.  You can customize
-set set of target CUDA architectues by setting `CMAKE_CUDA_ARCHITECTURES` (e.g. "50;60;70")
+libraries, and `CUDACXX` to the location of the nvcc compiler. You can customize
+a set of target CUDA architectures by setting `CMAKE_CUDA_ARCHITECTURES` (e.g. "50;60;70")
 
 Then generate dependencies:
 
@@ -67,15 +69,15 @@ go build .
 
 #### Linux ROCm (AMD)
 
-*Your operating system distribution may already have packages for AMD ROCm and CLBlast. Distro packages are often preferable, but instructions are distro-specific. Please consult distro-specific docs for dependencies if available!*
+_Your operating system distribution may already have packages for AMD ROCm and CLBlast. Distro packages are often preferable, but instructions are distro-specific. Please consult distro-specific docs for dependencies if available!_
 
-Install [CLBlast](https://github.com/CNugteren/CLBlast/blob/master/doc/installation.md) and [ROCm](https://rocm.docs.amd.com/en/latest/deploy/linux/quick_start.html) developement packages first, as well as `cmake` and `golang`.
+Install [CLBlast](https://github.com/CNugteren/CLBlast/blob/master/doc/installation.md) and [ROCm](https://rocm.docs.amd.com/en/latest/) development packages first, as well as `cmake` and `golang`.
 
 Typically the build scripts will auto-detect ROCm, however, if your Linux distro
 or installation approach uses unusual paths, you can specify the location by
 specifying an environment variable `ROCM_PATH` to the location of the ROCm
 install (typically `/opt/rocm`), and `CLBlast_DIR` to the location of the
-CLBlast install (typically `/usr/lib/cmake/CLBlast`).  You can also customize
+CLBlast install (typically `/usr/lib/cmake/CLBlast`). You can also customize
 the AMD GPU targets by setting AMDGPU_TARGETS (e.g. `AMDGPU_TARGETS="gfx1101;gfx1102"`)
 
 ```
@@ -88,51 +90,61 @@ Then build the binary:
 go build .
 ```
 
-ROCm requires elevated privileges to access the GPU at runtime.  On most distros you can add your user account to the `render` group, or run as root.
+ROCm requires elevated privileges to access the GPU at runtime. On most distros you can add your user account to the `render` group, or run as root.
 
 #### Advanced CPU Settings
 
 By default, running `go generate ./...` will compile a few different variations
 of the LLM library based on common CPU families and vector math capabilities,
 including a lowest-common-denominator which should run on almost any 64 bit CPU
-somewhat slowly.  At runtime, Ollama will auto-detect the optimal variation to
-load.  If you would like to build a CPU-based build customized for your
+somewhat slowly. At runtime, Ollama will auto-detect the optimal variation to
+load. If you would like to build a CPU-based build customized for your
 processor, you can set `OLLAMA_CUSTOM_CPU_DEFS` to the llama.cpp flags you would
-like to use.  For example, to compile an optimized binary for an Intel i9-9880H,
+like to use. For example, to compile an optimized binary for an Intel i9-9880H,
 you might use:
 
 ```
-OLLAMA_CUSTOM_CPU_DEFS="-DLLAMA_AVX=on -DLLAMA_AVX2=on -DLLAMA_F16C=on -DLLAMA_FMA=on" go generate ./...
+OLLAMA_CUSTOM_CPU_DEFS="-DGGML_AVX=on -DGGML_AVX2=on -DGGML_F16C=on -DGGML_FMA=on" go generate ./...
 go build .
 ```
 
 #### Containerized Linux Build
 
-If you have Docker available, you can build linux binaries with `./scripts/build_linux.sh` which has the CUDA and ROCm dependencies included.  The resulting binary is placed in `./dist`
-
+If you have Docker available, you can build linux binaries with `./scripts/build_linux.sh` which has the CUDA and ROCm dependencies included. The resulting binary is placed in `./dist`
 
 ### Windows
 
-Note: The windows build for Ollama is still under development.
+Note: The Windows build for Ollama is still under development.
 
-Install required tools:
+First, install required tools:
 
 - MSVC toolchain - C/C++ and cmake as minimal requirements
-- go version 1.21 or higher
+- Go version 1.22 or higher
 - MinGW (pick one variant) with GCC.
-  - <https://www.mingw-w64.org/>
-  - <https://www.msys2.org/>
+  - [MinGW-w64](https://www.mingw-w64.org/)
+  - [MSYS2](https://www.msys2.org/)
+- The `ThreadJob` Powershell module: `Install-Module -Name ThreadJob -Scope CurrentUser`
+
+Then, build the `ollama` binary:
 
 ```powershell
 $env:CGO_ENABLED="1"
-
 go generate ./...
-
 go build .
 ```
 
 #### Windows CUDA (NVIDIA)
 
-In addition to the common Windows development tools described above, install:
+In addition to the common Windows development tools described above, install CUDA after installing MSVC.
 
 - [NVIDIA CUDA](https://docs.nvidia.com/cuda/cuda-installation-guide-microsoft-windows/index.html)
+
+
+#### Windows ROCm (AMD Radeon)
+
+In addition to the common Windows development tools described above, install AMDs HIP package after installing MSVC.
+
+- [AMD HIP](https://www.amd.com/en/developer/resources/rocm-hub/hip-sdk.html)
+- [Strawberry Perl](https://strawberryperl.com/)
+
+Lastly, add `ninja.exe` included with MSVC to the system path (e.g. `C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\Common7\IDE\CommonExtensions\Microsoft\CMake\Ninja`).

docs/docker.md

@@ -0,0 +1,71 @@
+# Ollama Docker image
+
+### CPU only
+
+```bash
+docker run -d -v ollama:/root/.ollama -p 11434:11434 --name ollama ollama/ollama
+```
+
+### Nvidia GPU
+Install the [NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html#installation).
+
+#### Install with Apt
+1.  Configure the repository
+```bash
+curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey \
+    | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg
+curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list \
+    | sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' \
+    | sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list
+sudo apt-get update
+```
+2.  Install the NVIDIA Container Toolkit packages
+```bash
+sudo apt-get install -y nvidia-container-toolkit
+```
+
+#### Install with Yum or Dnf
+1.  Configure the repository
+    
+```bash
+curl -s -L https://nvidia.github.io/libnvidia-container/stable/rpm/nvidia-container-toolkit.repo \
+    | sudo tee /etc/yum.repos.d/nvidia-container-toolkit.repo
+```
+    
+2. Install the NVIDIA Container Toolkit packages
+    
+```bash
+sudo yum install -y nvidia-container-toolkit
+```
+
+#### Configure Docker to use Nvidia driver 
+```
+sudo nvidia-ctk runtime configure --runtime=docker
+sudo systemctl restart docker
+```
+
+#### Start the container
+
+```bash
+docker run -d --gpus=all -v ollama:/root/.ollama -p 11434:11434 --name ollama ollama/ollama
+```
+
+### AMD GPU
+
+To run Ollama using Docker with AMD GPUs, use the `rocm` tag and the following command:
+
+```
+docker run -d --device /dev/kfd --device /dev/dri -v ollama:/root/.ollama -p 11434:11434 --name ollama ollama/ollama:rocm
+```
+
+### Run model locally
+
+Now you can run a model:
+
+```
+docker exec -it ollama ollama run llama3
+```
+
+### Try different models
+
+More models can be found on the [Ollama library](https://ollama.com/library).

docs/faq.md

@@ -6,7 +6,7 @@ Ollama on macOS and Windows will automatically download updates. Click on the ta
 
 On Linux, re-run the install script:
 
-```
+```shell
 curl -fsSL https://ollama.com/install.sh | sh
 ```
 
@@ -14,6 +14,10 @@ curl -fsSL https://ollama.com/install.sh | sh
 
 Review the [Troubleshooting](./troubleshooting.md) docs for more about using logs.
 
+## Is my GPU compatible with Ollama?
+
+Please refer to the [GPU docs](./gpu.md).
+
 ## How can I specify the context window size?
 
 By default, Ollama uses a context window size of 2048 tokens.
@@ -26,9 +30,9 @@ To change this when using `ollama run`, use `/set parameter`:
 
 When using the API, specify the `num_ctx` parameter:
 
-```
+```shell
 curl http://localhost:11434/api/generate -d '{
-  "model": "llama2",
+  "model": "llama3",
   "prompt": "Why is the sky blue?",
   "options": {
     "num_ctx": 4096
@@ -36,6 +40,21 @@ curl http://localhost:11434/api/generate -d '{
 }'
 ```
 
+## How can I tell if my model was loaded onto the GPU?
+
+Use the `ollama ps` command to see what models are currently loaded into memory.
+
+```shell
+ollama ps
+NAME      	ID          	SIZE 	PROCESSOR	UNTIL
+llama3:70b	bcfb190ca3a7	42 GB	100% GPU 	4 minutes from now
+```
+
+The `Processor` column will show which memory the model was loaded in to:
+* `100% GPU` means the model was loaded entirely into the GPU
+* `100% CPU` means the model was loaded entirely in system memory
+* `48%/52% CPU/GPU` means the model was loaded partially onto both the GPU and into system memory
+
 ## How do I configure Ollama server?
 
 Ollama server can be configured with environment variables.
@@ -76,50 +95,19 @@ If Ollama is run as a systemd service, environment variables should be set using
 
 ### Setting environment variables on Windows
 
-On windows, Ollama inherits your user and system environment variables.
+On Windows, Ollama inherits your user and system environment variables.
 
-1. First Quit Ollama by clicking on it in the task bar
+1. First Quit Ollama by clicking on it in the task bar.
 
-2. Edit system environment variables from the control panel
+2. Start the Settings (Windows 11) or Control Panel (Windows 10) application and search for _environment variables_.
 
-3. Edit or create New variable(s) for your user account for `OLLAMA_HOST`, `OLLAMA_MODELS`, etc.
+3. Click on _Edit environment variables for your account_.
 
-4. Click OK/Apply to save 
+4. Edit or create a new variable for your user account for `OLLAMA_HOST`, `OLLAMA_MODELS`, etc.
 
-5. Run `ollama` from a new terminal window 
+5. Click OK/Apply to save.
 
-
-## How can I expose Ollama on my network?
-
-Ollama binds 127.0.0.1 port 11434 by default. Change the bind address with the `OLLAMA_HOST` environment variable.
-
-Refer to the section [above](#how-do-i-configure-ollama-server) for how to set environment variables on your platform.
-
-## How can I allow additional web origins to access Ollama?
-
-Ollama allows cross-origin requests from `127.0.0.1` and `0.0.0.0` by default. Additional origins can be configured with `OLLAMA_ORIGINS`.
-
-Refer to the section [above](#how-do-i-configure-ollama-server) for how to set environment variables on your platform.
-
-## Where are models stored?
-
-- macOS: `~/.ollama/models`
-- Linux: `/usr/share/ollama/.ollama/models`
-- Windows: `C:\Users\<username>\.ollama\models`
-
-### How do I set them to a different location?
-
-If a different directory needs to be used, set the environment variable `OLLAMA_MODELS` to the chosen directory.
-
-Refer to the section [above](#how-do-i-configure-ollama-server) for how to set environment variables on your platform.
-
-## Does Ollama send my prompts and answers back to Ollama.ai to use in any way?
-
-No, Ollama runs entirely locally, and conversation data will never leave your machine.
-
-## How can I use Ollama in Visual Studio Code?
-
-There is already a large collection of plugins available for VSCode as well as other editors that leverage Ollama. See the list of [extensions & plugins](https://github.com/jmorganca/ollama#extensions--plugins) at the bottom of the main repository readme.
+6. Start the Ollama application from the Windows Start menu.
 
 ## How do I use Ollama behind a proxy?
 
@@ -146,6 +134,69 @@ docker build -t ollama-with-ca .
 docker run -d -e HTTPS_PROXY=https://my.proxy.example.com -p 11434:11434 ollama-with-ca
 ```
 
+## Does Ollama send my prompts and answers back to ollama.com?
+
+No. Ollama runs locally, and conversation data does not leave your machine.
+
+## How can I expose Ollama on my network?
+
+Ollama binds 127.0.0.1 port 11434 by default. Change the bind address with the `OLLAMA_HOST` environment variable.
+
+Refer to the section [above](#how-do-i-configure-ollama-server) for how to set environment variables on your platform.
+
+## How can I use Ollama with a proxy server?
+
+Ollama runs an HTTP server and can be exposed using a proxy server such as Nginx. To do so, configure the proxy to forward requests and optionally set required headers (if not exposing Ollama on the network). For example, with Nginx:
+
+```
+server {
+    listen 80;
+    server_name example.com;  # Replace with your domain or IP
+    location / {
+        proxy_pass http://localhost:11434;
+        proxy_set_header Host localhost:11434;
+    }
+}
+```
+
+## How can I use Ollama with ngrok?
+
+Ollama can be accessed using a range of tools for tunneling tools. For example with Ngrok:
+
+```shell
+ngrok http 11434 --host-header="localhost:11434"
+```
+
+## How can I use Ollama with Cloudflare Tunnel?
+
+To use Ollama with Cloudflare Tunnel, use the `--url` and `--http-host-header` flags:
+
+```shell
+cloudflared tunnel --url http://localhost:11434 --http-host-header="localhost:11434"
+```
+
+## How can I allow additional web origins to access Ollama?
+
+Ollama allows cross-origin requests from `127.0.0.1` and `0.0.0.0` by default. Additional origins can be configured with `OLLAMA_ORIGINS`.
+
+Refer to the section [above](#how-do-i-configure-ollama-server) for how to set environment variables on your platform.
+
+## Where are models stored?
+
+- macOS: `~/.ollama/models`
+- Linux: `/usr/share/ollama/.ollama/models`
+- Windows: `C:\Users\%username%\.ollama\models`
+
+### How do I set them to a different location?
+
+If a different directory needs to be used, set the environment variable `OLLAMA_MODELS` to the chosen directory.
+
+Refer to the section [above](#how-do-i-configure-ollama-server) for how to set environment variables on your platform.
+
+## How can I use Ollama in Visual Studio Code?
+
+There is already a large collection of plugins available for VSCode as well as other editors that leverage Ollama. See the list of [extensions & plugins](https://github.com/ollama/ollama#extensions--plugins) at the bottom of the main repository readme.
+
 ## How do I use Ollama with GPU acceleration in Docker?
 
 The Ollama Docker container can be configured with GPU acceleration in Linux or Windows (with WSL2). This requires the [nvidia-container-toolkit](https://github.com/NVIDIA/nvidia-container-toolkit). See [ollama/ollama](https://hub.docker.com/r/ollama/ollama) for more details.
@@ -160,7 +211,7 @@ Open `Control Panel > Networking and Internet > View network status and tasks` a
 Click on `Configure` and open the `Advanced` tab. Search through each of the properties until you find `Large Send Offload Version 2 (IPv4)` and `Large Send Offload Version 2 (IPv6)`. *Disable* both of these
 properties.
 
-## How can I pre-load a model to get faster response times?
+## How can I preload a model into Ollama to get faster response times?
 
 If you are using the API you can preload a model by sending the Ollama server an empty request. This works with both the `/api/generate` and `/api/chat` API endpoints.
 
@@ -174,6 +225,11 @@ To use the chat completions endpoint, use:
 curl http://localhost:11434/api/chat -d '{"model": "mistral"}'
 ```
 
+To preload a model using the CLI, use the command:
+```shell
+ollama run llama3 ""
+```
+
 ## How do I keep a model loaded in memory or make it unload immediately?
 
 By default models are kept in memory for 5 minutes before being unloaded. This allows for quicker response times if you are making numerous requests to the LLM. You may, however, want to free up the memory before the 5 minutes have elapsed or keep the model loaded indefinitely. Use the `keep_alive` parameter with either the `/api/generate` and `/api/chat` API endpoints to control how long the model is left in memory.
@@ -186,10 +242,34 @@ The `keep_alive` parameter can be set to:
 
 For example, to preload a model and leave it in memory use:
 ```shell
-curl http://localhost:11434/api/generate -d '{"model": "llama2", "keep_alive": -1}'
+curl http://localhost:11434/api/generate -d '{"model": "llama3", "keep_alive": -1}'
 ```
 
 To unload the model and free up memory use:
 ```shell
-curl http://localhost:11434/api/generate -d '{"model": "llama2", "keep_alive": 0}'
+curl http://localhost:11434/api/generate -d '{"model": "llama3", "keep_alive": 0}'
 ```
+
+Alternatively, you can change the amount of time all models are loaded into memory by setting the `OLLAMA_KEEP_ALIVE` environment variable when starting the Ollama server. The `OLLAMA_KEEP_ALIVE` variable uses the same parameter types as the `keep_alive` parameter types mentioned above. Refer to section explaining [how to configure the Ollama server](#how-do-i-configure-ollama-server) to correctly set the environment variable.
+
+If you wish to override the `OLLAMA_KEEP_ALIVE` setting, use the `keep_alive` API parameter with the `/api/generate` or `/api/chat` API endpoints.
+
+## How do I manage the maximum number of requests the Ollama server can queue?
+
+If too many requests are sent to the server, it will respond with a 503 error indicating the server is overloaded.  You can adjust how many requests may be queue by setting `OLLAMA_MAX_QUEUE`.
+
+## How does Ollama handle concurrent requests?
+
+Ollama supports two levels of concurrent processing.  If your system has sufficient available memory (system memory when using CPU inference, or VRAM for GPU inference) then multiple models can be loaded at the same time.  For a given model, if there is sufficient available memory when the model is loaded, it is configured to allow parallel request processing.
+
+If there is insufficient available memory to load a new model request while one or more models are already loaded, all new requests will be queued until the new model can be loaded.  As prior models become idle, one or more will be unloaded to make room for the new model.  Queued requests will be processed in order.  When using GPU inference new models must be able to completely fit in VRAM to allow concurrent model loads.
+
+Parallel request processing for a given model results in increasing the context size by the number of parallel requests.  For example, a 2K context with 4 parallel requests will result in an 8K context and additional memory allocation.
+
+The following server settings may be used to adjust how Ollama handles concurrent requests on most platforms:
+
+- `OLLAMA_MAX_LOADED_MODELS` - The maximum number of models that can be loaded concurrently provided they fit in available memory.  The default is 3 * the number of GPUs or 3 for CPU inference.
+- `OLLAMA_NUM_PARALLEL` - The maximum number of parallel requests each model will process at the same time.  The default will auto-select either 4 or 1 based on available memory.
+- `OLLAMA_MAX_QUEUE` - The maximum number of requests Ollama will queue when busy before rejecting additional requests. The default is 512
+
+Note: Windows with Radeon GPUs currently default to 1 model maximum due to limitations in ROCm v5.7 for available VRAM reporting.  Once ROCm v6.2 is available, Windows Radeon will follow the defaults above.  You may enable concurrent model loads on Radeon on Windows, but ensure you don't load more models than will fit into your GPUs VRAM.

docs/gpu.md

@@ -0,0 +1,113 @@
+# GPU
+## Nvidia
+Ollama supports Nvidia GPUs with compute capability 5.0+.
+
+Check your compute compatibility to see if your card is supported:
+[https://developer.nvidia.com/cuda-gpus](https://developer.nvidia.com/cuda-gpus)
+
+| Compute Capability | Family              | Cards                                                                                                       |
+| ------------------ | ------------------- | ----------------------------------------------------------------------------------------------------------- |
+| 9.0                | NVIDIA              | `H100`                                                                                                      |
+| 8.9                | GeForce RTX 40xx    | `RTX 4090` `RTX 4080 SUPER` `RTX 4080` `RTX 4070 Ti SUPER` `RTX 4070 Ti` `RTX 4070 SUPER` `RTX 4070` `RTX 4060 Ti` `RTX 4060`  |
+|                    | NVIDIA Professional | `L4` `L40` `RTX 6000`                                                                                       |
+| 8.6                | GeForce RTX 30xx    | `RTX 3090 Ti` `RTX 3090` `RTX 3080 Ti` `RTX 3080` `RTX 3070 Ti` `RTX 3070` `RTX 3060 Ti` `RTX 3060`         |
+|                    | NVIDIA Professional | `A40` `RTX A6000` `RTX A5000` `RTX A4000` `RTX A3000` `RTX A2000` `A10` `A16` `A2`                          |
+| 8.0                | NVIDIA              | `A100` `A30`                                                                                                |
+| 7.5                | GeForce GTX/RTX     | `GTX 1650 Ti` `TITAN RTX` `RTX 2080 Ti` `RTX 2080` `RTX 2070` `RTX 2060`                                    |
+|                    | NVIDIA Professional | `T4` `RTX 5000` `RTX 4000` `RTX 3000` `T2000` `T1200` `T1000` `T600` `T500`                                 |
+|                    | Quadro              | `RTX 8000` `RTX 6000` `RTX 5000` `RTX 4000`                                                                 |
+| 7.0                | NVIDIA              | `TITAN V` `V100` `Quadro GV100`                                                                             |
+| 6.1                | NVIDIA TITAN        | `TITAN Xp` `TITAN X`                                                                                        |
+|                    | GeForce GTX         | `GTX 1080 Ti` `GTX 1080` `GTX 1070 Ti` `GTX 1070` `GTX 1060` `GTX 1050 Ti` `GTX 1050`                       |
+|                    | Quadro              | `P6000` `P5200` `P4200` `P3200` `P5000` `P4000` `P3000` `P2200` `P2000` `P1000` `P620` `P600` `P500` `P520` |
+|                    | Tesla               | `P40` `P4`                                                                                                  |
+| 6.0                | NVIDIA              | `Tesla P100` `Quadro GP100`                                                                                 |
+| 5.2                | GeForce GTX         | `GTX TITAN X` `GTX 980 Ti` `GTX 980` `GTX 970` `GTX 960` `GTX 950`                                          |
+|                    | Quadro              | `M6000 24GB` `M6000` `M5000` `M5500M` `M4000` `M2200` `M2000` `M620`                                        |
+|                    | Tesla               | `M60` `M40`                                                                                                 |
+| 5.0                | GeForce GTX         | `GTX 750 Ti` `GTX 750` `NVS 810`                                                                            |
+|                    | Quadro              | `K2200` `K1200` `K620` `M1200` `M520` `M5000M` `M4000M` `M3000M` `M2000M` `M1000M` `K620M` `M600M` `M500M`  |
+
+
+### GPU Selection
+
+If you have multiple NVIDIA GPUs in your system and want to limit Ollama to use
+a subset, you can set `CUDA_VISIBLE_DEVICES` to a comma separated list of GPUs.
+Numeric IDs may be used, however ordering may vary, so UUIDs are more reliable.
+You can discover the UUID of your GPUs by running `nvidia-smi -L` If you want to
+ignore the GPUs and force CPU usage, use an invalid GPU ID (e.g., "-1")
+
+### Laptop Suspend Resume
+
+On linux, after a suspend/resume cycle, sometimes Ollama will fail to discover
+your NVIDIA GPU, and fallback to running on the CPU.  You can workaround this
+driver bug by reloading the NVIDIA UVM driver with `sudo rmmod nvidia_uvm &&
+sudo modprobe nvidia_uvm`
+
+## AMD Radeon
+Ollama supports the following AMD GPUs:
+
+### Linux Support
+| Family         | Cards and accelerators                                                                                                               |
+| -------------- | ---------------------------------------------------------------------------------------------------------------------------------------------- |
+| AMD Radeon RX  | `7900 XTX` `7900 XT` `7900 GRE` `7800 XT` `7700 XT` `7600 XT` `7600` `6950 XT` `6900 XTX` `6900XT` `6800 XT` `6800` `Vega 64` `Vega 56`    |
+| AMD Radeon PRO | `W7900` `W7800` `W7700` `W7600` `W7500` `W6900X` `W6800X Duo` `W6800X` `W6800` `V620` `V420` `V340` `V320` `Vega II Duo` `Vega II` `VII` `SSG` |
+| AMD Instinct   | `MI300X` `MI300A` `MI300` `MI250X` `MI250` `MI210` `MI200` `MI100` `MI60` `MI50`                                                               |
+
+### Windows Support
+With ROCm v6.1, the following GPUs are supported on Windows.
+
+| Family         | Cards and accelerators                                                                                                               |
+| -------------- | ---------------------------------------------------------------------------------------------------------------------------------------------- |
+| AMD Radeon RX  | `7900 XTX` `7900 XT` `7900 GRE` `7800 XT` `7700 XT` `7600 XT` `7600` `6950 XT` `6900 XTX` `6900XT` `6800 XT` `6800`    |
+| AMD Radeon PRO | `W7900` `W7800` `W7700` `W7600` `W7500` `W6900X` `W6800X Duo` `W6800X` `W6800` `V620` |
+
+
+### Overrides on Linux
+Ollama leverages the AMD ROCm library, which does not support all AMD GPUs. In
+some cases you can force the system to try to use a similar LLVM target that is
+close.  For example The Radeon RX 5400 is `gfx1034` (also known as 10.3.4)
+however, ROCm does not currently support this target. The closest support is
+`gfx1030`.  You can use the environment variable `HSA_OVERRIDE_GFX_VERSION` with
+`x.y.z` syntax.  So for example, to force the system to run on the RX 5400, you
+would set `HSA_OVERRIDE_GFX_VERSION="10.3.0"` as an environment variable for the
+server.  If you have an unsupported AMD GPU you can experiment using the list of
+supported types below.
+
+At this time, the known supported GPU types on linux are the following LLVM Targets.
+This table shows some example GPUs that map to these LLVM targets:
+| **LLVM Target** | **An Example GPU** |
+|-----------------|---------------------|
+| gfx900 | Radeon RX Vega 56 |
+| gfx906 | Radeon Instinct MI50 |
+| gfx908 | Radeon Instinct MI100 |
+| gfx90a | Radeon Instinct MI210 |
+| gfx940 | Radeon Instinct MI300 |
+| gfx941 | |
+| gfx942 | |
+| gfx1030 | Radeon PRO V620 |
+| gfx1100 | Radeon PRO W7900 |
+| gfx1101 | Radeon PRO W7700 |
+| gfx1102 | Radeon RX 7600 |
+
+AMD is working on enhancing ROCm v6 to broaden support for families of GPUs in a
+future release which should increase support for more GPUs.
+
+Reach out on [Discord](https://discord.gg/ollama) or file an
+[issue](https://github.com/ollama/ollama/issues) for additional help.
+
+### GPU Selection
+
+If you have multiple AMD GPUs in your system and want to limit Ollama to use a
+subset, you can set `HIP_VISIBLE_DEVICES` to a comma separated list of GPUs.
+You can see the list of devices with `rocminfo`.  If you want to ignore the GPUs
+and force CPU usage, use an invalid GPU ID (e.g., "-1")
+
+### Container Permission
+
+In some Linux distributions, SELinux can prevent containers from
+accessing the AMD GPU devices.  On the host system you can run 
+`sudo setsebool container_use_devices=1` to allow containers to use devices.
+
+### Metal (Apple GPUs)
+Ollama supports GPU acceleration on Apple devices via the Metal API.

docs/import.md

@@ -1,165 +1,88 @@
-# Import a model
+# Import
 
-This guide walks through importing a GGUF, PyTorch or Safetensors model.
+GGUF models and select Safetensors models can be imported directly into Ollama.
 
-## Importing (GGUF)
+## Import GGUF
 
-### Step 1: Write a `Modelfile`
+A binary GGUF file can be imported directly into Ollama through a Modelfile.
 
-Start by creating a `Modelfile`. This file is the blueprint for your model, specifying weights, parameters, prompt templates and more.
-
-```
-FROM ./mistral-7b-v0.1.Q4_0.gguf
+```dockerfile
+FROM /path/to/file.gguf
 ```
 
-(Optional) many chat models require a prompt template in order to answer correctly. A default prompt template can be specified with the `TEMPLATE` instruction in the `Modelfile`:
+## Import Safetensors
 
-```
-FROM ./mistral-7b-v0.1.Q4_0.gguf
-TEMPLATE "[INST] {{ .Prompt }} [/INST]"
+If the model being imported is one of these architectures, it can be imported directly into Ollama through a Modelfile:
+
+ - LlamaForCausalLM
+ - MistralForCausalLM
+ - GemmaForCausalLM
+
+```dockerfile
+FROM /path/to/safetensors/directory
 ```
 
-### Step 2: Create the Ollama model
+For architectures not directly convertable by Ollama, see llama.cpp's [guide](https://github.com/ggerganov/llama.cpp/blob/master/README.md#prepare-and-quantize) on conversion. After conversion, see [Import GGUF](#import-gguf).
 
-Finally, create a model from your `Modelfile`:
+## Automatic Quantization
 
+> [!NOTE]
+> Automatic quantization requires v0.1.35 or higher.
+
+Ollama is capable of quantizing FP16 or FP32 models to any of the supported quantizations with the `-q/--quantize` flag in `ollama create`.
+
+```dockerfile
+FROM /path/to/my/gemma/f16/model
 ```
-ollama create example -f Modelfile
-```
-
-### Step 3: Run your model
-
-Next, test the model with `ollama run`:
-
-```
-ollama run example "What is your favourite condiment?"
-```
-
-## Importing (PyTorch & Safetensors)
-
-> Importing from PyTorch and Safetensors is a longer process than importing from GGUF. Improvements that make it easier are a work in progress.
-
-### Setup
-
-First, clone the `ollama/ollama` repo:
-
-```
-git clone git@github.com:ollama/ollama.git ollama
-cd ollama
-```
-
-and then fetch its `llama.cpp` submodule:
 
 ```shell
-git submodule init
-git submodule update llm/llama.cpp
+$ ollama create -q Q4_K_M mymodel
+transferring model data
+quantizing F16 model to Q4_K_M
+creating new layer sha256:735e246cc1abfd06e9cdcf95504d6789a6cd1ad7577108a70d9902fef503c1bd
+creating new layer sha256:0853f0ad24e5865173bbf9ffcc7b0f5d56b66fd690ab1009867e45e7d2c4db0f
+writing manifest
+success
 ```
 
-Next, install the Python dependencies:
+### Supported Quantizations
 
-```
-python3 -m venv llm/llama.cpp/.venv
-source llm/llama.cpp/.venv/bin/activate
-pip install -r llm/llama.cpp/requirements.txt
+- `Q4_0`
+- `Q4_1`
+- `Q5_0`
+- `Q5_1`
+- `Q8_0`
+
+#### K-means Quantizations
+
+- `Q3_K_S`
+- `Q3_K_M`
+- `Q3_K_L`
+- `Q4_K_S`
+- `Q4_K_M`
+- `Q5_K_S`
+- `Q5_K_M`
+- `Q6_K`
+
+## Template Detection
+
+> [!NOTE]
+> Template detection requires v0.1.42 or higher.
+
+Ollama uses model metadata, specifically `tokenizer.chat_template`, to automatically create a template appropriate for the model you're importing.
+
+```dockerfile
+FROM /path/to/my/gemma/model
 ```
 
-Then build the `quantize` tool:
-
-```
-make -C llm/llama.cpp quantize
+```shell
+$ ollama create mymodel
+transferring model data
+using autodetected template gemma-instruct
+creating new layer sha256:baa2a0edc27d19cc6b7537578a9a7ba1a4e3214dc185ed5ae43692b319af7b84
+creating new layer sha256:ba66c3309914dbef07e5149a648fd1877f030d337a4f240d444ea335008943cb
+writing manifest
+success
 ```
 
-### Clone the HuggingFace repository (optional)
-
-If the model is currently hosted in a HuggingFace repository, first clone that repository to download the raw model.
-
-Install [Git LFS](https://docs.github.com/en/repositories/working-with-files/managing-large-files/installing-git-large-file-storage), verify it's installed, and then clone the model's repository:
-
-```
-git lfs install
-git clone https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1 model
-```
-
-### Convert the model
-
-> Note: some model architectures require using specific convert scripts. For example, Qwen models require running `convert-hf-to-gguf.py` instead of `convert.py`
-
-```
-python llm/llama.cpp/convert.py ./model --outtype f16 --outfile converted.bin
-```
-
-### Quantize the model
-
-```
-llm/llama.cpp/quantize converted.bin quantized.bin q4_0
-```
-
-### Step 3: Write a `Modelfile`
-
-Next, create a `Modelfile` for your model:
-
-```
-FROM quantized.bin
-TEMPLATE "[INST] {{ .Prompt }} [/INST]"
-```
-
-### Step 4: Create the Ollama model
-
-Finally, create a model from your `Modelfile`:
-
-```
-ollama create example -f Modelfile
-```
-
-### Step 5: Run your model
-
-Next, test the model with `ollama run`:
-
-```
-ollama run example "What is your favourite condiment?"
-```
-
-## Publishing your model (optional – early alpha)
-
-Publishing models is in early alpha. If you'd like to publish your model to share with others, follow these steps:
-
-1. Create [an account](https://ollama.com/signup)
-2. Run `cat ~/.ollama/id_ed25519.pub` (or `type %USERPROFILE%\.ollama\id_ed25519.pub` on Windows) to view your Ollama public key. Copy this to the clipboard.
-3. Add your public key to your [Ollama account](https://ollama.com/settings/keys)
-
-Next, copy your model to your username's namespace:
-
-```
-ollama cp example <your username>/example
-```
-
-Then push the model:
-
-```
-ollama push <your username>/example
-```
-
-After publishing, your model will be available at `https://ollama.com/<your username>/example`.
-
-## Quantization reference
-
-The quantization options are as follow (from highest highest to lowest levels of quantization). Note: some architectures such as Falcon do not support K quants.
-
-- `q2_K`
-- `q3_K`
-- `q3_K_S`
-- `q3_K_M`
-- `q3_K_L`
-- `q4_0` (recommended)
-- `q4_1`
-- `q4_K`
-- `q4_K_S`
-- `q4_K_M`
-- `q5_0`
-- `q5_1`
-- `q5_K`
-- `q5_K_S`
-- `q5_K_M`
-- `q6_K`
-- `q8_0`
-- `f16`
+Defining a template in the Modelfile will disable this feature which may be useful if you want to use a different template than the autodetected one.

docs/linux.md

@@ -10,6 +10,14 @@ Install Ollama running this one-liner:
 curl -fsSL https://ollama.com/install.sh | sh
 ```
 
+## AMD Radeon GPU support
+
+While AMD has contributed the `amdgpu` driver upstream to the official linux
+kernel source, the version is older and may not support all ROCm features. We
+recommend you install the latest driver from
+https://www.amd.com/en/support/linux-drivers for best support of your Radeon
+GPU.
+
 ## Manual install
 
 ### Download the `ollama` binary
@@ -64,6 +72,11 @@ Verify that the drivers are installed by running the following command, which sh
 nvidia-smi
 ```
 
+### Install ROCm (optional - for Radeon GPUs)
+[Download and Install](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/tutorial/quick-start.html)
+
+Make sure to install ROCm v6
+
 ### Start Ollama
 
 Start Ollama using `systemd`:
@@ -87,12 +100,22 @@ sudo curl -L https://ollama.com/download/ollama-linux-amd64 -o /usr/bin/ollama
 sudo chmod +x /usr/bin/ollama
 ```
 
+## Installing specific versions
+
+Use `OLLAMA_VERSION` environment variable with the install script to install a specific version of Ollama, including pre-releases. You can find the version numbers in the [releases page](https://github.com/ollama/ollama/releases). 
+
+For example:
+
+```
+curl -fsSL https://ollama.com/install.sh | OLLAMA_VERSION=0.1.32 sh
+```
+
 ## Viewing logs
 
 To view logs of Ollama running as a startup service, run:
 
 ```bash
-journalctl -u ollama
+journalctl -e -u ollama
 ```
 
 ## Uninstall

docs/modelfile.md

@@ -10,7 +10,7 @@ A model file is the blueprint to create and share models with Ollama.
 - [Examples](#examples)
 - [Instructions](#instructions)
   - [FROM (Required)](#from-required)
-    - [Build from llama2](#build-from-llama2)
+    - [Build from llama3](#build-from-llama3)
     - [Build from a bin file](#build-from-a-bin-file)
   - [PARAMETER](#parameter)
     - [Valid Parameters and Values](#valid-parameters-and-values)
@@ -48,7 +48,7 @@ INSTRUCTION arguments
 An example of a `Modelfile` creating a mario blueprint:
 
 ```modelfile
-FROM llama2
+FROM llama3
 # sets the temperature to 1 [higher is more creative, lower is more coherent]
 PARAMETER temperature 1
 # sets the context window size to 4096, this controls how many tokens the LLM can use as context to generate the next token
@@ -67,33 +67,25 @@ To use this:
 
 More examples are available in the [examples directory](../examples).
 
-### `Modelfile`s in [ollama.com/library][1]
-
-There are two ways to view `Modelfile`s underlying the models in [ollama.com/library][1]:
-
-- Option 1: view a details page from a model's tags page:
-  1.  Go to a particular model's tags (e.g. https://ollama.com/library/llama2/tags)
-  2.  Click on a tag (e.g. https://ollama.com/library/llama2:13b)
-  3.  Scroll down to "Layers"
-      - Note: if the [`FROM` instruction](#from-required) is not present,
-        it means the model was created from a local file
-- Option 2: use `ollama show` to print the `Modelfile` for any local models like so:
+To view the Modelfile of a given model, use the `ollama show --modelfile` command.
 
   ```bash
-  > ollama show --modelfile llama2:13b
+  > ollama show --modelfile llama3
   # Modelfile generated by "ollama show"
   # To build a new Modelfile based on this one, replace the FROM line with:
-  # FROM llama2:13b
+  # FROM llama3:latest
+  FROM /Users/pdevine/.ollama/models/blobs/sha256-00e1317cbf74d901080d7100f57580ba8dd8de57203072dc6f668324ba545f29
+  TEMPLATE """{{ if .System }}<|start_header_id|>system<|end_header_id|>
 
-  FROM /root/.ollama/models/blobs/sha256:123abc
-  TEMPLATE """[INST] {{ if .System }}<<SYS>>{{ .System }}<</SYS>>
+  {{ .System }}<|eot_id|>{{ end }}{{ if .Prompt }}<|start_header_id|>user<|end_header_id|>
 
-  {{ end }}{{ .Prompt }} [/INST] """
-  SYSTEM """"""
-  PARAMETER stop [INST]
-  PARAMETER stop [/INST]
-  PARAMETER stop <<SYS>>
-  PARAMETER stop <</SYS>>
+  {{ .Prompt }}<|eot_id|>{{ end }}<|start_header_id|>assistant<|end_header_id|>
+
+  {{ .Response }}<|eot_id|>"""
+  PARAMETER stop "<|start_header_id|>"
+  PARAMETER stop "<|end_header_id|>"
+  PARAMETER stop "<|eot_id|>"
+  PARAMETER stop "<|reserved_special_token"
   ```
 
 ## Instructions
@@ -106,14 +98,14 @@ The `FROM` instruction defines the base model to use when creating a model.
 FROM <model name>:<tag>
 ```
 
-#### Build from llama2
+#### Build from llama3
 
 ```modelfile
-FROM llama2
+FROM llama3
 ```
 
 A list of available base models:
-<https://github.com/jmorganca/ollama#model-library>
+<https://github.com/ollama/ollama#model-library>
 
 #### Build from a `bin` file
 
@@ -131,7 +123,7 @@ The `PARAMETER` instruction defines a parameter that can be set when the model i
 PARAMETER <parameter> <parametervalue>
 ```
 
-### Valid Parameters and Values
+#### Valid Parameters and Values
 
 | Parameter      | Description                                                                                                                                                                                                                                             | Value Type | Example Usage        |
 | -------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------- | -------------------- |
@@ -139,9 +131,6 @@ PARAMETER <parameter> <parametervalue>
 | mirostat_eta   | Influences how quickly the algorithm responds to feedback from the generated text. A lower learning rate will result in slower adjustments, while a higher learning rate will make the algorithm more responsive. (Default: 0.1)                        | float      | mirostat_eta 0.1     |
 | mirostat_tau   | Controls the balance between coherence and diversity of the output. A lower value will result in more focused and coherent text. (Default: 5.0)                                                                                                         | float      | mirostat_tau 5.0     |
 | num_ctx        | Sets the size of the context window used to generate the next token. (Default: 2048)                                                                                                                                                                    | int        | num_ctx 4096         |
-| num_gqa        | The number of GQA groups in the transformer layer. Required for some models, for example it is 8 for llama2:70b                                                                                                                                         | int        | num_gqa 1            |
-| num_gpu        | The number of layers to send to the GPU(s). On macOS it defaults to 1 to enable metal support, 0 to disable.                                                                                                                                            | int        | num_gpu 50           |
-| num_thread     | Sets the number of threads to use during computation. By default, Ollama will detect this for optimal performance. It is recommended to set this value to the number of physical CPU cores your system has (as opposed to the logical number of cores). | int        | num_thread 8         |
 | repeat_last_n  | Sets how far back for the model to look back to prevent repetition. (Default: 64, 0 = disabled, -1 = num_ctx)                                                                                                                                           | int        | repeat_last_n 64     |
 | repeat_penalty | Sets how strongly to penalize repetitions. A higher value (e.g., 1.5) will penalize repetitions more strongly, while a lower value (e.g., 0.9) will be more lenient. (Default: 1.1)                                                                     | float      | repeat_penalty 1.1   |
 | temperature    | The temperature of the model. Increasing the temperature will make the model answer more creatively. (Default: 0.8)                                                                                                                                     | float      | temperature 0.7      |
@@ -183,7 +172,7 @@ SYSTEM """<system message>"""
 
 ### ADAPTER
 
-The `ADAPTER` instruction specifies the LoRA adapter to apply to the base model. The value of this instruction should be an absolute path or a path relative to the Modelfile and the file must be in a GGML file format. The adapter should be tuned from the base model otherwise the behaviour is undefined.
+The `ADAPTER` instruction is an optional instruction that specifies any LoRA adapter that should apply to the base model. The value of this instruction should be an absolute path or a path relative to the Modelfile and the file must be in a GGML file format. The adapter should be tuned from the base model otherwise the behaviour is undefined.
 
 ```modelfile
 ADAPTER ./ollama-lora.bin
@@ -201,7 +190,22 @@ LICENSE """
 
 ### MESSAGE
 
-The `MESSAGE` instruction allows you to specify a message history for the model to use when responding:
+The `MESSAGE` instruction allows you to specify a message history for the model to use when responding. Use multiple iterations of the MESSAGE command to build up a conversation which will guide the model to answer in a similar way.
+
+```modelfile
+MESSAGE <role> <message>
+```
+
+#### Valid roles
+
+| Role      | Description                                                  |
+| --------- | ------------------------------------------------------------ |
+| system    | Alternate way of providing the SYSTEM message for the model. |
+| user      | An example message of what the user could have asked.        |
+| assistant | An example message of how the model should respond.          |
+
+
+#### Example conversation
 
 ```modelfile
 MESSAGE user Is Toronto in Canada?
@@ -212,6 +216,7 @@ MESSAGE user Is Ontario in Canada?
 MESSAGE assistant yes
 ```
 
+
 ## Notes
 
 - the **`Modelfile` is not case sensitive**. In the examples, uppercase instructions are used to make it easier to distinguish it from arguments.

docs/openai.md

@@ -1,6 +1,6 @@
 # OpenAI compatibility
 
-> **Note:** OpenAI compatibility is experimental and is subject to major adjustments including breaking changes. For fully-featured access to the Ollama API, see the Ollama [Python library](https://github.com/ollama/ollama-python), [JavaScript library](https://github.com/ollama/ollama-js) and [REST API](https://github.com/jmorganca/ollama/blob/main/docs/api.md).
+> **Note:** OpenAI compatibility is experimental and is subject to major adjustments including breaking changes. For fully-featured access to the Ollama API, see the Ollama [Python library](https://github.com/ollama/ollama-python), [JavaScript library](https://github.com/ollama/ollama-js) and [REST API](https://github.com/ollama/ollama/blob/main/docs/api.md).
 
 Ollama provides experimental compatibility with parts of the [OpenAI API](https://platform.openai.com/docs/api-reference) to help connect existing applications to Ollama.
 
@@ -25,7 +25,7 @@ chat_completion = client.chat.completions.create(
             'content': 'Say this is a test',
         }
     ],
-    model='llama2',
+    model='llama3',
 )
 ```
 
@@ -43,7 +43,7 @@ const openai = new OpenAI({
 
 const chatCompletion = await openai.chat.completions.create({
   messages: [{ role: 'user', content: 'Say this is a test' }],
-  model: 'llama2',
+  model: 'llama3',
 })
 ```
 
@@ -53,7 +53,7 @@ const chatCompletion = await openai.chat.completions.create({
 curl http://localhost:11434/v1/chat/completions \
     -H "Content-Type: application/json" \
     -d '{
-        "model": "llama2",
+        "model": "llama3",
         "messages": [
             {
                 "role": "system",
@@ -65,6 +65,7 @@ curl http://localhost:11434/v1/chat/completions \
             }
         ]
     }'
+
 ```
 
 ## Endpoints
@@ -102,18 +103,12 @@ curl http://localhost:11434/v1/chat/completions \
 - [ ] `user`
 - [ ] `n`
 
-#### Notes
-
-- Setting `seed` will always set `temperature` to `0`
-- `finish_reason` will always be `stop`
-- `usage.prompt_tokens` will be 0 for completions where prompt evaluation is cached
-
 ## Models
 
 Before using a model, pull it locally `ollama pull`:
 
 ```shell
-ollama pull llama2
+ollama pull llama3
 ```
 
 ### Default model names
@@ -121,7 +116,7 @@ ollama pull llama2
 For tooling that relies on default OpenAI model names such as `gpt-3.5-turbo`, use `ollama cp` to copy an existing model name to a temporary name:
 
 ```
-ollama cp llama2 gpt-3.5-turbo
+ollama cp llama3 gpt-3.5-turbo
 ```
 
 Afterwards, this new model name can be specified the `model` field:

docs/troubleshooting.md

@@ -1,72 +1,96 @@
-# How to troubleshoot issues
-
-Sometimes Ollama may not perform as expected. One of the best ways to figure out what happened is to take a look at the logs. Find the logs on **Mac** by running the command:
-
-```shell
-cat ~/.ollama/logs/server.log
-```
-
-On **Linux** systems with systemd, the logs can be found with this command:
-
-```shell
-journalctl -u ollama
-```
-
-When you run Ollama in a **container**, the logs go to stdout/stderr in the container:
-
-```shell
-docker logs <container-name>
-```
-(Use `docker ps` to find the container name)
-
-If manually running `ollama serve` in a terminal, the logs will be on that terminal.
-
-When you run Ollama on **Windows**, there are a few different locations.  You can view them in the explorer window by hitting `<cmd>+R` and type in:
-- `explorer %LOCALAPPDATA%\Ollama` to view logs
-- `explorer %LOCALAPPDATA%\Programs\Ollama` to browse the binaries (The installer adds this to your user PATH)
-- `explorer %HOMEPATH%\.ollama` to browse where models and configuration is stored
-- `explorer %TEMP%` where temporary executable files are stored in one or more `ollama*` directories
-
-To enable additional debug logging to help troubleshoot problems, first **Quit the running app from the tray menu** then in a powershell terminal
-```powershell
-$env:OLLAMA_DEBUG="1"
-& "ollama app.exe"
-```
-
-Join the [Discord](https://discord.gg/ollama) for help interpreting the logs.
-
-## LLM libraries
-
-Ollama includes multiple LLM libraries compiled for different GPUs and CPU
-vector features.  Ollama tries to pick the best one based on the capabilities of
-your system.  If this autodetection has problems, or you run into other problems
-(e.g. crashes in your GPU) you can workaround this by forcing a specific LLM
-library.  `cpu_avx2` will perform the best, followed by `cpu_avx` an the slowest
-but most compatible is `cpu`.  Rosetta emulation under MacOS will work with the
-`cpu` library. 
-
-In the server log, you will see a message that looks something like this (varies
-from release to release):
-
-```
-Dynamic LLM libraries [rocm_v6 cpu cpu_avx cpu_avx2 cuda_v11 rocm_v5]
-```
-
-**Experimental LLM Library Override**
-
-You can set OLLAMA_LLM_LIBRARY to any of the available LLM libraries to bypass
-autodetection, so for example, if you have a CUDA card, but want to force the
-CPU LLM library with AVX2 vector support, use:
-
-```
-OLLAMA_LLM_LIBRARY="cpu_avx2" ollama serve
-```
-
-You can see what features your CPU has with the following.  
-```
-cat /proc/cpuinfo| grep flags  | head -1
-```
-
-## Known issues
-
-* N/A
+# How to troubleshoot issues
+
+Sometimes Ollama may not perform as expected. One of the best ways to figure out what happened is to take a look at the logs. Find the logs on **Mac** by running the command:
+
+```shell
+cat ~/.ollama/logs/server.log
+```
+
+On **Linux** systems with systemd, the logs can be found with this command:
+
+```shell
+journalctl -u ollama
+```
+
+When you run Ollama in a **container**, the logs go to stdout/stderr in the container:
+
+```shell
+docker logs <container-name>
+```
+(Use `docker ps` to find the container name)
+
+If manually running `ollama serve` in a terminal, the logs will be on that terminal.
+
+When you run Ollama on **Windows**, there are a few different locations. You can view them in the explorer window by hitting `<cmd>+R` and type in:
+- `explorer %LOCALAPPDATA%\Ollama` to view logs.  The most recent server logs will be in `server.log` and older logs will be in `server-#.log` 
+- `explorer %LOCALAPPDATA%\Programs\Ollama` to browse the binaries (The installer adds this to your user PATH)
+- `explorer %HOMEPATH%\.ollama` to browse where models and configuration is stored
+- `explorer %TEMP%` where temporary executable files are stored in one or more `ollama*` directories
+
+To enable additional debug logging to help troubleshoot problems, first **Quit the running app from the tray menu** then in a powershell terminal
+```powershell
+$env:OLLAMA_DEBUG="1"
+& "ollama app.exe"
+```
+
+Join the [Discord](https://discord.gg/ollama) for help interpreting the logs.
+
+## LLM libraries
+
+Ollama includes multiple LLM libraries compiled for different GPUs and CPU vector features. Ollama tries to pick the best one based on the capabilities of your system. If this autodetection has problems, or you run into other problems (e.g. crashes in your GPU) you can workaround this by forcing a specific LLM library. `cpu_avx2` will perform the best, followed by `cpu_avx` an the slowest but most compatible is `cpu`. Rosetta emulation under MacOS will work with the `cpu` library. 
+
+In the server log, you will see a message that looks something like this (varies from release to release):
+
+```
+Dynamic LLM libraries [rocm_v6 cpu cpu_avx cpu_avx2 cuda_v11 rocm_v5]
+```
+
+**Experimental LLM Library Override**
+
+You can set OLLAMA_LLM_LIBRARY to any of the available LLM libraries to bypass autodetection, so for example, if you have a CUDA card, but want to force the CPU LLM library with AVX2 vector support, use:
+
+```
+OLLAMA_LLM_LIBRARY="cpu_avx2" ollama serve
+```
+
+You can see what features your CPU has with the following.
+```
+cat /proc/cpuinfo| grep flags | head -1
+```
+
+## Installing older or pre-release versions on Linux
+
+If you run into problems on Linux and want to install an older version, or you'd like to try out a pre-release before it's officially released, you can tell the install script which version to install.
+
+```sh
+curl -fsSL https://ollama.com/install.sh | OLLAMA_VERSION="0.1.29" sh
+```
+
+## Linux tmp noexec 
+
+If your system is configured with the "noexec" flag where Ollama stores its temporary executable files, you can specify an alternate location by setting OLLAMA_TMPDIR to a location writable by the user ollama runs as. For example OLLAMA_TMPDIR=/usr/share/ollama/
+
+## NVIDIA GPU Discovery
+
+When Ollama starts up, it takes inventory of the GPUs present in the system to determine compatibility and how much VRAM is available.  Sometimes this discovery can fail to find your GPUs.  In general, running the latest driver will yield the best results.
+
+### Linux NVIDIA Troubleshooting
+
+If you are using a container to run Ollama, make sure you've set up the container runtime first as described in [docker.md](./docker.md)
+
+Sometimes the Ollama can have difficulties initializing the GPU. When you check the server logs, this can show up as various error codes, such as "3" (not initialized), "46" (device unavailable), "100" (no device), "999" (unknown), or others. The following troubleshooting techniques may help resolve the problem
+
+- If you are using a container, is the container runtime working?  Try `docker run --gpus all ubuntu nvidia-smi` - if this doesn't work, Ollama wont be able to see your NVIDIA GPU.
+- Is the uvm driver loaded? `sudo nvidia-modprobe -u`
+- Try reloading the nvidia_uvm driver - `sudo rmmod nvidia_uvm` then `sudo modprobe nvidia_uvm`
+- Try rebooting
+- Make sure you're running the latest nvidia drivers
+
+If none of those resolve the problem, gather additional information and file an issue:
+- Set `CUDA_ERROR_LEVEL=50` and try again to get more diagnostic logs
+- Check dmesg for any errors `sudo dmesg | grep -i nvrm` and `sudo dmesg | grep -i nvidia`
+
+
+## Windows Terminal Errors
+
+Older versions of Windows 10 (e.g., 21H1) are known to have a bug where the standard terminal program does not display control characters correctly.  This can result in a long string of strings like `←[?25h←[?25l` being displayed, sometimes erroring with `The parameter is incorrect`  To resolve this problem, please update to Win 10 22H1 or newer.

docs/tutorials/langchainjs.md

@@ -5,28 +5,28 @@ In this tutorial, we are going to use JavaScript with LangChain and Ollama to le
 To get started, let's just use **LangChain** to ask a simple question to a model. To do this with JavaScript, we need to install **LangChain**:
 
 ```bash
-npm install langchain
+npm install @langchain/community
 ```
 
 Now we can start building out our JavaScript:
 
 ```javascript
-import { Ollama } from "langchain/llms/ollama";
+import { Ollama } from "@langchain/community/llms/ollama";
 
 const ollama = new Ollama({
   baseUrl: "http://localhost:11434",
-  model: "llama2",
+  model: "llama3",
 });
 
-const answer = await ollama.call(`why is the sky blue?`);
+const answer = await ollama.invoke(`why is the sky blue?`);
 
 console.log(answer);
 ```
 
-That will get us the same thing as if we ran `ollama run llama2 "why is the sky blue"` in the terminal. But we want to load a document from the web to ask a question against. **Cheerio** is a great library for ingesting a webpage, and **LangChain** uses it in their **CheerioWebBaseLoader**. So let's install **Cheerio** and build that part of the app.
+That will get us the same thing as if we ran `ollama run llama3 "why is the sky blue"` in the terminal. But we want to load a document from the web to ask a question against. **Cheerio** is a great library for ingesting a webpage, and **LangChain** uses it in their **CheerioWebBaseLoader**. So let's install **Cheerio** and build that part of the app.
 
 ```bash
-npm install cheerio 
+npm install cheerio
 ```
 
 ```javascript

docs/tutorials/langchainpy.md

@@ -12,15 +12,17 @@ So let's figure out how we can use **LangChain** with Ollama to ask our question
 
 Let's start by asking a simple question that we can get an answer to from the **Llama2** model using **Ollama**. First, we need to install the **LangChain** package:
 
-`pip install langchain`
+`pip install langchain_community`
 
 Then we can create a model and ask the question:
 
 ```python
-from langchain.llms import Ollama
-ollama = Ollama(base_url='http://localhost:11434',
-model="llama2")
-print(ollama("why is the sky blue"))
+from langchain_community.llms import Ollama
+ollama = Ollama(
+    base_url='http://localhost:11434',
+    model="llama3"
+)
+print(ollama.invoke("why is the sky blue"))
 ```
 
 Notice that we are defining the model and the base URL for Ollama.
@@ -42,12 +44,12 @@ text_splitter=RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=0)
 all_splits = text_splitter.split_documents(data)
 ```
 
-It's split up, but we have to find the relevant splits and then submit those to the model. We can do this by creating embeddings and storing them in a vector database. We can use Ollama directly to instantiate an embedding model. We will use ChromaDB in this example for a vector database. `pip install GPT4All chromadb`
-
+It's split up, but we have to find the relevant splits and then submit those to the model. We can do this by creating embeddings and storing them in a vector database. We can use Ollama directly to instantiate an embedding model. We will use ChromaDB in this example for a vector database. `pip install chromadb`
+We also need to pull embedding model: `ollama pull nomic-embed-text`
 ```python
 from langchain.embeddings import OllamaEmbeddings
 from langchain.vectorstores import Chroma
-oembed = OllamaEmbeddings(base_url="http://localhost:11434", model="llama2")
+oembed = OllamaEmbeddings(base_url="http://localhost:11434", model="nomic-embed-text")
 vectorstore = Chroma.from_documents(documents=all_splits, embedding=oembed)
 ```
 
@@ -66,7 +68,8 @@ The next thing is to send the question and the relevant parts of the docs to the
 ```python
 from langchain.chains import RetrievalQA
 qachain=RetrievalQA.from_chain_type(ollama, retriever=vectorstore.as_retriever())
-qachain({"query": question})
+res = qachain.invoke({"query": question})
+print(res['result'])
 ```
 
 The answer received from this chain was:

docs/tutorials/nvidia-jetson.md

@@ -1,38 +1,15 @@
 # Running Ollama on NVIDIA Jetson Devices
 
-With some minor configuration, Ollama runs well on [NVIDIA Jetson Devices](https://www.nvidia.com/en-us/autonomous-machines/embedded-systems/). The following has been tested on [JetPack 5.1.2](https://developer.nvidia.com/embedded/jetpack).
+Ollama runs well on [NVIDIA Jetson Devices](https://www.nvidia.com/en-us/autonomous-machines/embedded-systems/) and should run out of the box with the standard installation instructions. 
 
-NVIDIA Jetson devices are Linux-based embedded AI computers that are purpose-built for AI applications.
-
-Jetsons have an integrated GPU that is wired directly to the memory controller of the machine. For this reason, the `nvidia-smi` command is unrecognized, and Ollama proceeds to operate in "CPU only"
-mode. This can be verified by using a monitoring tool like jtop.
-
-In order to address this, we simply pass the path to the Jetson's pre-installed CUDA libraries into `ollama serve` (while in a tmux session). We then hardcode the num_gpu parameters into a cloned
-version of our target model.
-
-Prerequisites:
-
-- curl
-- tmux
-
-Here are the steps:
+The following has been tested on [JetPack 5.1.2](https://developer.nvidia.com/embedded/jetpack), but should also work on JetPack 6.0.
 
 - Install Ollama via standard Linux command (ignore the 404 error): `curl https://ollama.com/install.sh | sh`
-- Stop the Ollama service: `sudo systemctl stop ollama`
-- Start Ollama serve in a tmux session called ollama_jetson and reference the CUDA libraries path: `tmux has-session -t ollama_jetson 2>/dev/null || tmux new-session -d -s ollama_jetson 
-'LD_LIBRARY_PATH=/usr/local/cuda/lib64 ollama serve'`
 - Pull the model you want to use (e.g. mistral): `ollama pull mistral`
-- Create a new Modelfile specifically for enabling GPU support on the Jetson: `touch ModelfileMistralJetson`
-- In the ModelfileMistralJetson file, specify the FROM model and the num_gpu PARAMETER as shown below:
-
-```
-FROM mistral
-PARAMETER num_gpu 999
-```
-
-- Create a new model from your Modelfile: `ollama create mistral-jetson -f ./ModelfileMistralJetson`
-- Run the new model: `ollama run mistral-jetson`
-
-If you run a monitoring tool like jtop you should now see that Ollama is using the Jetson's integrated GPU.
+- Start an interactive session: `ollama run mistral`
 
 And that's it!
+
+# Running Ollama in Docker
+
+When running GPU accelerated applications in Docker, it is highly recommended to use [dusty-nv jetson-containers repo](https://github.com/dusty-nv/jetson-containers).

docs/windows.md

@@ -1,46 +1,61 @@
-# Ollama Windows Preview
-
-Welcome to the Ollama Windows preview.
-
-No more WSL required!
-
-Ollama now runs as a native Windows application, including NVIDIA GPU support.
-After installing Ollama Windows Preview, Ollama will run in the background and
-the `ollama` command line is available in `cmd`, `powershell` or your favorite
-terminal application. As usual the Ollama [api](./api.md) will be served on
-`http://localhost:11434`.
-
-As this is a preview release, you should expect a few bugs here and there.  If
-you run into a problem you can reach out on
-[Discord](https://discord.gg/ollama), or file an 
-[issue](https://github.com/ollama/ollama/issues).
-Logs will often be helpful in dianosing the problem (see
-[Troubleshooting](#troubleshooting) below)
-
-## System Requirements
-
-* Windows 10 or newer, Home or Pro
-* NVIDIA 452.39 or newer Drivers if you have an NVIDIA card
-
-## API Access
-
-Here's a quick example showing API access from `powershell`
-```powershell
-(Invoke-WebRequest -method POST -Body '{"model":"llama2", "prompt":"Why is the sky blue?", "stream": false}' -uri http://localhost:11434/api/generate ).Content | ConvertFrom-json
-```
-
-## Troubleshooting
-
-While we're in preview, `OLLAMA_DEBUG` is always enabled, which adds
-a "view logs" menu item to the app, and increses logging for the GUI app and
-server.
-
-Ollama on Windows stores files in a few different locations.  You can view them in
-the explorer window by hitting `<cmd>+R` and type in:
-- `explorer %LOCALAPPDATA%\Ollama` contains logs, and downloaded updates
-    - *app.log* contains logs from the GUI application
-    - *server.log* contains the server logs
-    - *upgrade.log* contains log output for upgrades
-- `explorer %LOCALAPPDATA%\Programs\Ollama` contains the binaries (The installer adds this to your user PATH)
-- `explorer %HOMEPATH%\.ollama` contains models and configuration
-- `explorer %TEMP%` contains temporary executable files in one or more `ollama*` directories
+# Ollama Windows Preview
+
+Welcome to the Ollama Windows preview.
+
+No more WSL required!
+
+Ollama now runs as a native Windows application, including NVIDIA and AMD Radeon GPU support.
+After installing Ollama Windows Preview, Ollama will run in the background and
+the `ollama` command line is available in `cmd`, `powershell` or your favorite
+terminal application. As usual the Ollama [api](./api.md) will be served on
+`http://localhost:11434`.
+
+As this is a preview release, you should expect a few bugs here and there.  If
+you run into a problem you can reach out on
+[Discord](https://discord.gg/ollama), or file an
+[issue](https://github.com/ollama/ollama/issues).
+Logs will often be helpful in diagnosing the problem (see
+[Troubleshooting](#troubleshooting) below)
+
+## System Requirements
+
+* Windows 10 22H2 or newer, Home or Pro
+* NVIDIA 452.39 or newer Drivers if you have an NVIDIA card
+* AMD Radeon Driver https://www.amd.com/en/support if you have a Radeon card
+
+## API Access
+
+Here's a quick example showing API access from `powershell`
+```powershell
+(Invoke-WebRequest -method POST -Body '{"model":"llama3", "prompt":"Why is the sky blue?", "stream": false}' -uri http://localhost:11434/api/generate ).Content | ConvertFrom-json
+```
+
+## Troubleshooting
+
+While we're in preview, `OLLAMA_DEBUG` is always enabled, which adds
+a "view logs" menu item to the app, and increases logging for the GUI app and
+server.
+
+Ollama on Windows stores files in a few different locations.  You can view them in
+the explorer window by hitting `<cmd>+R` and type in:
+- `explorer %LOCALAPPDATA%\Ollama` contains logs, and downloaded updates
+    - *app.log* contains most resent logs from the GUI application
+    - *server.log* contains the most recent server logs
+    - *upgrade.log* contains log output for upgrades
+- `explorer %LOCALAPPDATA%\Programs\Ollama` contains the binaries (The installer adds this to your user PATH)
+- `explorer %HOMEPATH%\.ollama` contains models and configuration
+- `explorer %TEMP%` contains temporary executable files in one or more `ollama*` directories
+
+
+## Standalone CLI
+
+The easiest way to install Ollama on Windows is to use the `OllamaSetup.exe`
+installer. It installs in your account without requiring Administrator rights.
+We update Ollama regularly to support the latest models, and this installer will
+help you keep up to date.
+
+If you'd like to install or integrate Ollama as a service, a standalone
+`ollama-windows-amd64.zip` zip file is available containing only the Ollama CLI
+and GPU library dependencies for Nvidia and AMD. This allows for embedding
+Ollama in existing applications, or running it as a system service via `ollama
+serve` with tools such as [NSSM](https://nssm.cc/).

envconfig/config.go

@@ -0,0 +1,373 @@
+package envconfig
+
+import (
+	"errors"
+	"fmt"
+	"log/slog"
+	"math"
+	"net"
+	"os"
+	"path/filepath"
+	"runtime"
+	"strconv"
+	"strings"
+	"time"
+)
+
+type OllamaHost struct {
+	Scheme string
+	Host   string
+	Port   string
+}
+
+func (o OllamaHost) String() string {
+	return fmt.Sprintf("%s://%s:%s", o.Scheme, o.Host, o.Port)
+}
+
+var ErrInvalidHostPort = errors.New("invalid port specified in OLLAMA_HOST")
+
+var (
+	// Set via OLLAMA_ORIGINS in the environment
+	AllowOrigins []string
+	// Set via OLLAMA_DEBUG in the environment
+	Debug bool
+	// Experimental flash attention
+	FlashAttention bool
+	// Set via OLLAMA_HOST in the environment
+	Host *OllamaHost
+	// Set via OLLAMA_KEEP_ALIVE in the environment
+	KeepAlive time.Duration
+	// Set via OLLAMA_LLM_LIBRARY in the environment
+	LLMLibrary string
+	// Set via OLLAMA_MAX_LOADED_MODELS in the environment
+	MaxRunners int
+	// Set via OLLAMA_MAX_QUEUE in the environment
+	MaxQueuedRequests int
+	// Set via OLLAMA_MAX_VRAM in the environment
+	MaxVRAM uint64
+	// Set via OLLAMA_MODELS in the environment
+	ModelsDir string
+	// Set via OLLAMA_NOHISTORY in the environment
+	NoHistory bool
+	// Set via OLLAMA_NOPRUNE in the environment
+	NoPrune bool
+	// Set via OLLAMA_NUM_PARALLEL in the environment
+	NumParallel int
+	// Set via OLLAMA_RUNNERS_DIR in the environment
+	RunnersDir string
+	// Set via OLLAMA_SCHED_SPREAD in the environment
+	SchedSpread bool
+	// Set via OLLAMA_TMPDIR in the environment
+	TmpDir string
+	// Set via OLLAMA_INTEL_GPU in the environment
+	IntelGpu bool
+
+	// Set via CUDA_VISIBLE_DEVICES in the environment
+	CudaVisibleDevices string
+	// Set via HIP_VISIBLE_DEVICES in the environment
+	HipVisibleDevices string
+	// Set via ROCR_VISIBLE_DEVICES in the environment
+	RocrVisibleDevices string
+	// Set via GPU_DEVICE_ORDINAL in the environment
+	GpuDeviceOrdinal string
+	// Set via HSA_OVERRIDE_GFX_VERSION in the environment
+	HsaOverrideGfxVersion string
+)
+
+type EnvVar struct {
+	Name        string
+	Value       any
+	Description string
+}
+
+func AsMap() map[string]EnvVar {
+	ret := map[string]EnvVar{
+		"OLLAMA_DEBUG":             {"OLLAMA_DEBUG", Debug, "Show additional debug information (e.g. OLLAMA_DEBUG=1)"},
+		"OLLAMA_FLASH_ATTENTION":   {"OLLAMA_FLASH_ATTENTION", FlashAttention, "Enabled flash attention"},
+		"OLLAMA_HOST":              {"OLLAMA_HOST", Host, "IP Address for the ollama server (default 127.0.0.1:11434)"},
+		"OLLAMA_KEEP_ALIVE":        {"OLLAMA_KEEP_ALIVE", KeepAlive, "The duration that models stay loaded in memory (default \"5m\")"},
+		"OLLAMA_LLM_LIBRARY":       {"OLLAMA_LLM_LIBRARY", LLMLibrary, "Set LLM library to bypass autodetection"},
+		"OLLAMA_MAX_LOADED_MODELS": {"OLLAMA_MAX_LOADED_MODELS", MaxRunners, "Maximum number of loaded models per GPU"},
+		"OLLAMA_MAX_QUEUE":         {"OLLAMA_MAX_QUEUE", MaxQueuedRequests, "Maximum number of queued requests"},
+		"OLLAMA_MAX_VRAM":          {"OLLAMA_MAX_VRAM", MaxVRAM, "Maximum VRAM"},
+		"OLLAMA_MODELS":            {"OLLAMA_MODELS", ModelsDir, "The path to the models directory"},
+		"OLLAMA_NOHISTORY":         {"OLLAMA_NOHISTORY", NoHistory, "Do not preserve readline history"},
+		"OLLAMA_NOPRUNE":           {"OLLAMA_NOPRUNE", NoPrune, "Do not prune model blobs on startup"},
+		"OLLAMA_NUM_PARALLEL":      {"OLLAMA_NUM_PARALLEL", NumParallel, "Maximum number of parallel requests"},
+		"OLLAMA_ORIGINS":           {"OLLAMA_ORIGINS", AllowOrigins, "A comma separated list of allowed origins"},
+		"OLLAMA_RUNNERS_DIR":       {"OLLAMA_RUNNERS_DIR", RunnersDir, "Location for runners"},
+		"OLLAMA_SCHED_SPREAD":      {"OLLAMA_SCHED_SPREAD", SchedSpread, "Always schedule model across all GPUs"},
+		"OLLAMA_TMPDIR":            {"OLLAMA_TMPDIR", TmpDir, "Location for temporary files"},
+	}
+	if runtime.GOOS != "darwin" {
+		ret["CUDA_VISIBLE_DEVICES"] = EnvVar{"CUDA_VISIBLE_DEVICES", CudaVisibleDevices, "Set which NVIDIA devices are visible"}
+		ret["HIP_VISIBLE_DEVICES"] = EnvVar{"HIP_VISIBLE_DEVICES", HipVisibleDevices, "Set which AMD devices are visible"}
+		ret["ROCR_VISIBLE_DEVICES"] = EnvVar{"ROCR_VISIBLE_DEVICES", RocrVisibleDevices, "Set which AMD devices are visible"}
+		ret["GPU_DEVICE_ORDINAL"] = EnvVar{"GPU_DEVICE_ORDINAL", GpuDeviceOrdinal, "Set which AMD devices are visible"}
+		ret["HSA_OVERRIDE_GFX_VERSION"] = EnvVar{"HSA_OVERRIDE_GFX_VERSION", HsaOverrideGfxVersion, "Override the gfx used for all detected AMD GPUs"}
+		ret["OLLAMA_INTEL_GPU"] = EnvVar{"OLLAMA_INTEL_GPU", IntelGpu, "Enable experimental Intel GPU detection"}
+	}
+	return ret
+}
+
+func Values() map[string]string {
+	vals := make(map[string]string)
+	for k, v := range AsMap() {
+		vals[k] = fmt.Sprintf("%v", v.Value)
+	}
+	return vals
+}
+
+var defaultAllowOrigins = []string{
+	"localhost",
+	"127.0.0.1",
+	"0.0.0.0",
+}
+
+// Clean quotes and spaces from the value
+func clean(key string) string {
+	return strings.Trim(os.Getenv(key), "\"' ")
+}
+
+func init() {
+	// default values
+	NumParallel = 0 // Autoselect
+	MaxRunners = 0  // Autoselect
+	MaxQueuedRequests = 512
+	KeepAlive = 5 * time.Minute
+
+	LoadConfig()
+}
+
+func LoadConfig() {
+	if debug := clean("OLLAMA_DEBUG"); debug != "" {
+		d, err := strconv.ParseBool(debug)
+		if err == nil {
+			Debug = d
+		} else {
+			Debug = true
+		}
+	}
+
+	if fa := clean("OLLAMA_FLASH_ATTENTION"); fa != "" {
+		d, err := strconv.ParseBool(fa)
+		if err == nil {
+			FlashAttention = d
+		}
+	}
+
+	RunnersDir = clean("OLLAMA_RUNNERS_DIR")
+	if runtime.GOOS == "windows" && RunnersDir == "" {
+		// On Windows we do not carry the payloads inside the main executable
+		appExe, err := os.Executable()
+		if err != nil {
+			slog.Error("failed to lookup executable path", "error", err)
+		}
+
+		cwd, err := os.Getwd()
+		if err != nil {
+			slog.Error("failed to lookup working directory", "error", err)
+		}
+
+		var paths []string
+		for _, root := range []string{filepath.Dir(appExe), cwd} {
+			paths = append(paths,
+				root,
+				filepath.Join(root, "windows-"+runtime.GOARCH),
+				filepath.Join(root, "dist", "windows-"+runtime.GOARCH),
+			)
+		}
+
+		// Try a few variations to improve developer experience when building from source in the local tree
+		for _, p := range paths {
+			candidate := filepath.Join(p, "ollama_runners")
+			_, err := os.Stat(candidate)
+			if err == nil {
+				RunnersDir = candidate
+				break
+			}
+		}
+		if RunnersDir == "" {
+			slog.Error("unable to locate llm runner directory.  Set OLLAMA_RUNNERS_DIR to the location of 'ollama_runners'")
+		}
+	}
+
+	TmpDir = clean("OLLAMA_TMPDIR")
+
+	userLimit := clean("OLLAMA_MAX_VRAM")
+	if userLimit != "" {
+		avail, err := strconv.ParseUint(userLimit, 10, 64)
+		if err != nil {
+			slog.Error("invalid setting, ignoring", "OLLAMA_MAX_VRAM", userLimit, "error", err)
+		} else {
+			MaxVRAM = avail
+		}
+	}
+
+	LLMLibrary = clean("OLLAMA_LLM_LIBRARY")
+
+	if onp := clean("OLLAMA_NUM_PARALLEL"); onp != "" {
+		val, err := strconv.Atoi(onp)
+		if err != nil {
+			slog.Error("invalid setting, ignoring", "OLLAMA_NUM_PARALLEL", onp, "error", err)
+		} else {
+			NumParallel = val
+		}
+	}
+
+	if nohistory := clean("OLLAMA_NOHISTORY"); nohistory != "" {
+		NoHistory = true
+	}
+
+	if spread := clean("OLLAMA_SCHED_SPREAD"); spread != "" {
+		s, err := strconv.ParseBool(spread)
+		if err == nil {
+			SchedSpread = s
+		} else {
+			SchedSpread = true
+		}
+	}
+
+	if noprune := clean("OLLAMA_NOPRUNE"); noprune != "" {
+		NoPrune = true
+	}
+
+	if origins := clean("OLLAMA_ORIGINS"); origins != "" {
+		AllowOrigins = strings.Split(origins, ",")
+	}
+	for _, allowOrigin := range defaultAllowOrigins {
+		AllowOrigins = append(AllowOrigins,
+			fmt.Sprintf("http://%s", allowOrigin),
+			fmt.Sprintf("https://%s", allowOrigin),
+			fmt.Sprintf("http://%s", net.JoinHostPort(allowOrigin, "*")),
+			fmt.Sprintf("https://%s", net.JoinHostPort(allowOrigin, "*")),
+		)
+	}
+
+	AllowOrigins = append(AllowOrigins,
+		"app://*",
+		"file://*",
+		"tauri://*",
+	)
+
+	maxRunners := clean("OLLAMA_MAX_LOADED_MODELS")
+	if maxRunners != "" {
+		m, err := strconv.Atoi(maxRunners)
+		if err != nil {
+			slog.Error("invalid setting, ignoring", "OLLAMA_MAX_LOADED_MODELS", maxRunners, "error", err)
+		} else {
+			MaxRunners = m
+		}
+	}
+
+	if onp := os.Getenv("OLLAMA_MAX_QUEUE"); onp != "" {
+		p, err := strconv.Atoi(onp)
+		if err != nil || p <= 0 {
+			slog.Error("invalid setting, ignoring", "OLLAMA_MAX_QUEUE", onp, "error", err)
+		} else {
+			MaxQueuedRequests = p
+		}
+	}
+
+	ka := clean("OLLAMA_KEEP_ALIVE")
+	if ka != "" {
+		loadKeepAlive(ka)
+	}
+
+	var err error
+	ModelsDir, err = getModelsDir()
+	if err != nil {
+		slog.Error("invalid setting", "OLLAMA_MODELS", ModelsDir, "error", err)
+	}
+
+	Host, err = getOllamaHost()
+	if err != nil {
+		slog.Error("invalid setting", "OLLAMA_HOST", Host, "error", err, "using default port", Host.Port)
+	}
+
+	if set, err := strconv.ParseBool(clean("OLLAMA_INTEL_GPU")); err == nil {
+		IntelGpu = set
+	}
+
+	CudaVisibleDevices = clean("CUDA_VISIBLE_DEVICES")
+	HipVisibleDevices = clean("HIP_VISIBLE_DEVICES")
+	RocrVisibleDevices = clean("ROCR_VISIBLE_DEVICES")
+	GpuDeviceOrdinal = clean("GPU_DEVICE_ORDINAL")
+	HsaOverrideGfxVersion = clean("HSA_OVERRIDE_GFX_VERSION")
+}
+
+func getModelsDir() (string, error) {
+	if models, exists := os.LookupEnv("OLLAMA_MODELS"); exists {
+		return models, nil
+	}
+	home, err := os.UserHomeDir()
+	if err != nil {
+		return "", err
+	}
+	return filepath.Join(home, ".ollama", "models"), nil
+}
+
+func getOllamaHost() (*OllamaHost, error) {
+	defaultPort := "11434"
+
+	hostVar := os.Getenv("OLLAMA_HOST")
+	hostVar = strings.TrimSpace(strings.Trim(strings.TrimSpace(hostVar), "\"'"))
+
+	scheme, hostport, ok := strings.Cut(hostVar, "://")
+	switch {
+	case !ok:
+		scheme, hostport = "http", hostVar
+	case scheme == "http":
+		defaultPort = "80"
+	case scheme == "https":
+		defaultPort = "443"
+	}
+
+	// trim trailing slashes
+	hostport = strings.TrimRight(hostport, "/")
+
+	host, port, err := net.SplitHostPort(hostport)
+	if err != nil {
+		host, port = "127.0.0.1", defaultPort
+		if ip := net.ParseIP(strings.Trim(hostport, "[]")); ip != nil {
+			host = ip.String()
+		} else if hostport != "" {
+			host = hostport
+		}
+	}
+
+	if portNum, err := strconv.ParseInt(port, 10, 32); err != nil || portNum > 65535 || portNum < 0 {
+		return &OllamaHost{
+			Scheme: scheme,
+			Host:   host,
+			Port:   defaultPort,
+		}, ErrInvalidHostPort
+	}
+
+	return &OllamaHost{
+		Scheme: scheme,
+		Host:   host,
+		Port:   port,
+	}, nil
+}
+
+func loadKeepAlive(ka string) {
+	v, err := strconv.Atoi(ka)
+	if err != nil {
+		d, err := time.ParseDuration(ka)
+		if err == nil {
+			if d < 0 {
+				KeepAlive = time.Duration(math.MaxInt64)
+			} else {
+				KeepAlive = d
+			}
+		}
+	} else {
+		d := time.Duration(v) * time.Second
+		if d < 0 {
+			KeepAlive = time.Duration(math.MaxInt64)
+		} else {
+			KeepAlive = d
+		}
+	}
+}

envconfig/config_test.go

@@ -0,0 +1,88 @@
+package envconfig
+
+import (
+	"fmt"
+	"math"
+	"net"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestConfig(t *testing.T) {
+	Debug = false // Reset whatever was loaded in init()
+	t.Setenv("OLLAMA_DEBUG", "")
+	LoadConfig()
+	require.False(t, Debug)
+	t.Setenv("OLLAMA_DEBUG", "false")
+	LoadConfig()
+	require.False(t, Debug)
+	t.Setenv("OLLAMA_DEBUG", "1")
+	LoadConfig()
+	require.True(t, Debug)
+	t.Setenv("OLLAMA_FLASH_ATTENTION", "1")
+	LoadConfig()
+	require.True(t, FlashAttention)
+	t.Setenv("OLLAMA_KEEP_ALIVE", "")
+	LoadConfig()
+	require.Equal(t, 5*time.Minute, KeepAlive)
+	t.Setenv("OLLAMA_KEEP_ALIVE", "3")
+	LoadConfig()
+	require.Equal(t, 3*time.Second, KeepAlive)
+	t.Setenv("OLLAMA_KEEP_ALIVE", "1h")
+	LoadConfig()
+	require.Equal(t, 1*time.Hour, KeepAlive)
+	t.Setenv("OLLAMA_KEEP_ALIVE", "-1s")
+	LoadConfig()
+	require.Equal(t, time.Duration(math.MaxInt64), KeepAlive)
+	t.Setenv("OLLAMA_KEEP_ALIVE", "-1")
+	LoadConfig()
+	require.Equal(t, time.Duration(math.MaxInt64), KeepAlive)
+}
+
+func TestClientFromEnvironment(t *testing.T) {
+	type testCase struct {
+		value  string
+		expect string
+		err    error
+	}
+
+	hostTestCases := map[string]*testCase{
+		"empty":               {value: "", expect: "127.0.0.1:11434"},
+		"only address":        {value: "1.2.3.4", expect: "1.2.3.4:11434"},
+		"only port":           {value: ":1234", expect: ":1234"},
+		"address and port":    {value: "1.2.3.4:1234", expect: "1.2.3.4:1234"},
+		"hostname":            {value: "example.com", expect: "example.com:11434"},
+		"hostname and port":   {value: "example.com:1234", expect: "example.com:1234"},
+		"zero port":           {value: ":0", expect: ":0"},
+		"too large port":      {value: ":66000", err: ErrInvalidHostPort},
+		"too small port":      {value: ":-1", err: ErrInvalidHostPort},
+		"ipv6 localhost":      {value: "[::1]", expect: "[::1]:11434"},
+		"ipv6 world open":     {value: "[::]", expect: "[::]:11434"},
+		"ipv6 no brackets":    {value: "::1", expect: "[::1]:11434"},
+		"ipv6 + port":         {value: "[::1]:1337", expect: "[::1]:1337"},
+		"extra space":         {value: " 1.2.3.4 ", expect: "1.2.3.4:11434"},
+		"extra quotes":        {value: "\"1.2.3.4\"", expect: "1.2.3.4:11434"},
+		"extra space+quotes":  {value: " \" 1.2.3.4 \" ", expect: "1.2.3.4:11434"},
+		"extra single quotes": {value: "'1.2.3.4'", expect: "1.2.3.4:11434"},
+	}
+
+	for k, v := range hostTestCases {
+		t.Run(k, func(t *testing.T) {
+			t.Setenv("OLLAMA_HOST", v.value)
+			LoadConfig()
+
+			oh, err := getOllamaHost()
+			if err != v.err {
+				t.Fatalf("expected %s, got %s", v.err, err)
+			}
+
+			if err == nil {
+				host := net.JoinHostPort(oh.Host, oh.Port)
+				assert.Equal(t, v.expect, host, fmt.Sprintf("%s: expected %s, got %s", k, v.expect, host))
+			}
+		})
+	}
+}

examples/bash-comparemodels/README.md

@@ -1,10 +0,0 @@
-# Bash Shell examples
-
-When calling `ollama`, you can pass it a file to run all the prompts in the file, one after the other:
-
-`ollama run llama2 < sourcequestions.txt`
-
-This concept is used in the following example.
-
-## Compare Models
-`comparemodels.sh` is a script that runs all the questions in `sourcequestions.txt` using any 4 models you choose that you have already pulled from the Ollama library or have created locally.

examples/bash-comparemodels/comparemodels.sh

@@ -1,64 +0,0 @@
-#! /usr/bin/env bash
-# Compare multiple models by running them with the same questions
-
-NUMBEROFCHOICES=4
-SELECTIONS=()
-declare -a SUMS=()
-
-# Get the list of models
-CHOICES=$(ollama list | awk '{print $1}')
-
-# Select which models to run as a comparison
-echo "Select $NUMBEROFCHOICES models to compare:"
-select ITEM in $CHOICES; do
-    if [[ -n $ITEM ]]; then
-        echo "You have selected $ITEM"
-        SELECTIONS+=("$ITEM")
-        ((COUNT++))
-        if [[ $COUNT -eq $NUMBEROFCHOICES ]]; then
-            break
-        fi
-    else
-        echo "Invalid selection"
-    fi
-done
-
-# Loop through each of the selected models
-for ITEM in "${SELECTIONS[@]}"; do
-    echo "--------------------------------------------------------------"
-    echo "Loading the model $ITEM into memory"
-    ollama run "$ITEM" ""
-    echo "--------------------------------------------------------------"
-    echo "Running the questions through the model $ITEM"
-    COMMAND_OUTPUT=$(ollama run "$ITEM" --verbose < sourcequestions.txt 2>&1| tee /dev/stderr)
-
-    # eval duration is sometimes listed in seconds and sometimes in milliseconds. 
-    # Add up the values for each model
-    SUM=$(echo "$COMMAND_OUTPUT" | awk '
-    /eval duration:/ {
-        value = $3
-        if (index(value, "ms") > 0) {
-            gsub("ms", "", value)
-            value /= 1000
-        } else {
-            gsub("s", "", value)
-        }
-        sum += value
-    }
-    END { print sum }')
-
-
-    SUMS+=("All questions for $ITEM completed in $SUM seconds")
-done
-
-echo ""
-echo "--------------------------------------------------------------"
-echo -e "Sums of eval durations for each run:"
-for val in "${SUMS[@]}"; do
-    echo "$val"
-done
-
-echo "--------------------------------------------------------------"
-echo "Comparison complete. Now you can decide"
-echo "which model is best."
-echo "--------------------------------------------------------------"

examples/bash-comparemodels/sourcequestions.txt

@@ -1,7 +0,0 @@
-Why is the sky blue
-What is a black hole
-Explain the big bang theory like I am 5?
-What is the quickest way to win a game of Monopoly with 3 others?
-Why does a vacuum bottle keep my coffee hot and my milkshake cold?
-What is the difference between a meteor, a meteorite, and a meteoroid?
-Create an array with 5 items and print to the console. Do this in Python, C#, Typescript, and Rust.

examples/flyio/.gitignore

@@ -0,0 +1 @@
+fly.toml

examples/flyio/README.md

@@ -0,0 +1,67 @@
+# Deploy Ollama to Fly.io
+
+> Note: this example exposes a public endpoint and does not configure authentication. Use with care.
+
+## Prerequisites
+
+- Ollama: https://ollama.com/download
+- Fly.io account. Sign up for a free account: https://fly.io/app/sign-up
+
+## Steps
+
+1. Login to Fly.io
+
+    ```bash
+    fly auth login
+    ```
+
+1. Create a new Fly app
+
+    ```bash
+    fly launch --name <name> --image ollama/ollama --internal-port 11434 --vm-size shared-cpu-8x --now
+    ```
+
+1. Pull and run `orca-mini:3b`
+
+    ```bash
+    OLLAMA_HOST=https://<name>.fly.dev ollama run orca-mini:3b
+    ```
+
+`shared-cpu-8x` is a free-tier eligible machine type. For better performance, switch to a `performance` or `dedicated` machine type or attach a GPU for hardware acceleration (see below).
+
+## (Optional) Persistent Volume
+
+By default Fly Machines use ephemeral storage which is problematic if you want to use the same model across restarts without pulling it again. Create and attach a persistent volume to store the downloaded models:
+
+1. Create the Fly Volume
+
+    ```bash
+    fly volume create ollama
+    ```
+
+1. Update `fly.toml` and add `[mounts]`
+
+    ```toml
+    [mounts]
+      source = "ollama"
+      destination = "/mnt/ollama/models"
+    ```
+
+1. Update `fly.toml` and add `[env]`
+
+    ```toml
+    [env]
+      OLLAMA_MODELS = "/mnt/ollama/models"
+    ```
+
+1. Deploy your app
+
+    ```bash
+    fly deploy
+    ```
+
+## (Optional) Hardware Acceleration
+
+Fly.io GPU is currently in waitlist. Sign up for the waitlist: https://fly.io/gpu
+
+Once you've been accepted, create the app with the additional flags `--vm-gpu-kind a100-pcie-40gb` or `--vm-gpu-kind a100-pcie-80gb`.

examples/go-chat/main.go

@@ -0,0 +1,51 @@
+package main
+
+import (
+	"context"
+	"fmt"
+	"log"
+
+	"github.com/ollama/ollama/api"
+)
+
+func main() {
+	client, err := api.ClientFromEnvironment()
+	if err != nil {
+		log.Fatal(err)
+	}
+
+	messages := []api.Message{
+		api.Message{
+			Role:    "system",
+			Content: "Provide very brief, concise responses",
+		},
+		api.Message{
+			Role:    "user",
+			Content: "Name some unusual animals",
+		},
+		api.Message{
+			Role:    "assistant",
+			Content: "Monotreme, platypus, echidna",
+		},
+		api.Message{
+			Role:    "user",
+			Content: "which of these is the most dangerous?",
+		},
+	}
+
+	ctx := context.Background()
+	req := &api.ChatRequest{
+		Model:    "llama3",
+		Messages: messages,
+	}
+
+	respFunc := func(resp api.ChatResponse) error {
+		fmt.Print(resp.Message.Content)
+		return nil
+	}
+
+	err = client.Chat(ctx, req, respFunc)
+	if err != nil {
+		log.Fatal(err)
+	}
+}

examples/go-generate-streaming/main.go

@@ -0,0 +1,40 @@
+package main
+
+import (
+	"context"
+	"fmt"
+	"log"
+
+	"github.com/ollama/ollama/api"
+)
+
+func main() {
+	client, err := api.ClientFromEnvironment()
+	if err != nil {
+		log.Fatal(err)
+	}
+
+	// By default, GenerateRequest is streaming.
+	req := &api.GenerateRequest{
+		Model:  "gemma",
+		Prompt: "how many planets are there?",
+	}
+
+	ctx := context.Background()
+	respFunc := func(resp api.GenerateResponse) error {
+		// Only print the response here; GenerateResponse has a number of other
+		// interesting fields you want to examine.
+
+		// In streaming mode, responses are partial so we call fmt.Print (and not
+		// Println) in order to avoid spurious newlines being introduced. The
+		// model will insert its own newlines if it wants.
+		fmt.Print(resp.Response)
+		return nil
+	}
+
+	err = client.Generate(ctx, req, respFunc)
+	if err != nil {
+		log.Fatal(err)
+	}
+	fmt.Println()
+}

examples/go-generate/main.go

@@ -0,0 +1,37 @@
+package main
+
+import (
+	"context"
+	"fmt"
+	"log"
+
+	"github.com/ollama/ollama/api"
+)
+
+func main() {
+	client, err := api.ClientFromEnvironment()
+	if err != nil {
+		log.Fatal(err)
+	}
+
+	req := &api.GenerateRequest{
+		Model:  "gemma",
+		Prompt: "how many planets are there?",
+
+		// set streaming to false
+		Stream: new(bool),
+	}
+
+	ctx := context.Background()
+	respFunc := func(resp api.GenerateResponse) error {
+		// Only print the response here; GenerateResponse has a number of other
+		// interesting fields you want to examine.
+		fmt.Println(resp.Response)
+		return nil
+	}
+
+	err = client.Generate(ctx, req, respFunc)
+	if err != nil {
+		log.Fatal(err)
+	}
+}

examples/go-http-generate/main.go

@@ -19,7 +19,7 @@ func main() {
 	}
 
 	defer resp.Body.Close()
-	
+
 	responseData, err := io.ReadAll(resp.Body)
 	if err != nil {
 		log.Fatal(err)

examples/go-multimodal/main.go

@@ -0,0 +1,47 @@
+package main
+
+import (
+	"context"
+	"fmt"
+	"log"
+	"os"
+
+	"github.com/ollama/ollama/api"
+)
+
+func main() {
+	if len(os.Args) <= 1 {
+		log.Fatal("usage: <image name>")
+	}
+
+	imgData, err := os.ReadFile(os.Args[1])
+	if err != nil {
+		log.Fatal(err)
+	}
+
+	client, err := api.ClientFromEnvironment()
+	if err != nil {
+		log.Fatal(err)
+	}
+
+	req := &api.GenerateRequest{
+		Model:  "llava",
+		Prompt: "describe this image",
+		Images: []api.ImageData{imgData},
+	}
+
+	ctx := context.Background()
+	respFunc := func(resp api.GenerateResponse) error {
+		// In streaming mode, responses are partial so we call fmt.Print (and not
+		// Println) in order to avoid spurious newlines being introduced. The
+		// model will insert its own newlines if it wants.
+		fmt.Print(resp.Response)
+		return nil
+	}
+
+	err = client.Generate(ctx, req, respFunc)
+	if err != nil {
+		log.Fatal(err)
+	}
+	fmt.Println()
+}

examples/go-pull-progress/main.go

@@ -0,0 +1,31 @@
+package main
+
+import (
+	"context"
+	"fmt"
+	"log"
+
+	"github.com/ollama/ollama/api"
+)
+
+func main() {
+	client, err := api.ClientFromEnvironment()
+	if err != nil {
+		log.Fatal(err)
+	}
+
+	ctx := context.Background()
+
+	req := &api.PullRequest{
+		Model: "mistral",
+	}
+	progressFunc := func(resp api.ProgressResponse) error {
+		fmt.Printf("Progress: status=%v, total=%v, completed=%v\n", resp.Status, resp.Total, resp.Completed)
+		return nil
+	}
+
+	err = client.Pull(ctx, req, progressFunc)
+	if err != nil {
+		log.Fatal(err)
+	}
+}

examples/kubernetes/README.md

@@ -7,12 +7,24 @@
 
 ## Steps
 
-1. Create the Ollama namespace, daemon set, and service
+1. Create the Ollama namespace, deployment, and service
 
    ```bash
    kubectl apply -f cpu.yaml
    ```
 
+## (Optional) Hardware Acceleration
+
+Hardware acceleration in Kubernetes requires NVIDIA's [`k8s-device-plugin`](https://github.com/NVIDIA/k8s-device-plugin) which is deployed in Kubernetes in form of daemonset. Follow the link for more details.
+
+Once configured, create a GPU enabled Ollama deployment.
+
+```bash
+kubectl apply -f gpu.yaml
+```
+
+## Test
+
 1. Port forward the Ollama service to connect and use it locally
 
    ```bash
@@ -23,14 +35,4 @@
 
    ```bash
    ollama run orca-mini:3b
-   ```
-
-## (Optional) Hardware Acceleration
-
-Hardware acceleration in Kubernetes requires NVIDIA's [`k8s-device-plugin`](https://github.com/NVIDIA/k8s-device-plugin). Follow the link for more details.
-
-Once configured, create a GPU enabled Ollama deployment.
-
-```bash
-kubectl apply -f gpu.yaml
-```
+   ```

examples/langchain-python-rag-document/main.py

@@ -40,9 +40,9 @@ while True:
         continue
 
     # Prompt
-    template = """Use the following pieces of context to answer the question at the end. 
-    If you don't know the answer, just say that you don't know, don't try to make up an answer. 
-    Use three sentences maximum and keep the answer as concise as possible. 
+    template = """Use the following pieces of context to answer the question at the end.
+    If you don't know the answer, just say that you don't know, don't try to make up an answer.
+    Use three sentences maximum and keep the answer as concise as possible.
     {context}
     Question: {question}
     Helpful Answer:"""
@@ -51,11 +51,11 @@ while True:
         template=template,
     )
 
-    llm = Ollama(model="llama2:13b", callback_manager=CallbackManager([StreamingStdOutCallbackHandler()]))
+    llm = Ollama(model="llama3:8b", callback_manager=CallbackManager([StreamingStdOutCallbackHandler()]))
     qa_chain = RetrievalQA.from_chain_type(
         llm,
         retriever=vectorstore.as_retriever(),
         chain_type_kwargs={"prompt": QA_CHAIN_PROMPT},
     )
 
-    result = qa_chain({"query": query})
+    result = qa_chain({"query": query})

examples/langchain-python-rag-privategpt/README.md

@@ -1,6 +1,6 @@
 # PrivateGPT with Llama 2 uncensored
 
-https://github.com/jmorganca/ollama/assets/3325447/20cf8ec6-ff25-42c6-bdd8-9be594e3ce1b
+https://github.com/ollama/ollama/assets/3325447/20cf8ec6-ff25-42c6-bdd8-9be594e3ce1b
 
 > Note: this example is a slightly modified version of PrivateGPT using models such as Llama 2 Uncensored. All credit for PrivateGPT goes to Iván Martínez who is the creator of it, and you can find his GitHub repo [here](https://github.com/imartinez/privateGPT).
 

examples/langchain-python-rag-privategpt/ingest.py

@@ -77,13 +77,21 @@ LOADER_MAPPING = {
 
 
 def load_single_document(file_path: str) -> List[Document]:
-    ext = "." + file_path.rsplit(".", 1)[-1]
-    if ext in LOADER_MAPPING:
-        loader_class, loader_args = LOADER_MAPPING[ext]
-        loader = loader_class(file_path, **loader_args)
-        return loader.load()
+    if os.path.getsize(file_path) != 0:
+        filename, ext = os.path.splitext(file_path)
+        if ext in LOADER_MAPPING:
+            loader_class, loader_args = LOADER_MAPPING[ext]
+            try:
+                loader = loader_class(file_path, **loader_args)
+                if loader:
+                    return loader.load()
+            except:
+                print(f"Corrupted file {file_path}. Ignoring it.")
+        else:
+            print(f"Unsupported file {file_path}. Ignoring it.")
+    else:
+        print(f"Empty file {file_path}. Ignoring it.")
 
-    raise ValueError(f"Unsupported file extension '{ext}'")
 
 def load_documents(source_dir: str, ignored_files: List[str] = []) -> List[Document]:
     """
@@ -100,7 +108,8 @@ def load_documents(source_dir: str, ignored_files: List[str] = []) -> List[Docum
         results = []
         with tqdm(total=len(filtered_files), desc='Loading new documents', ncols=80) as pbar:
             for i, docs in enumerate(pool.imap_unordered(load_single_document, filtered_files)):
-                results.extend(docs)
+                if docs:
+                    results.extend(docs)
                 pbar.update()
 
     return results

examples/langchain-python-rag-privategpt/requirements.txt

@@ -11,4 +11,5 @@ tabulate==0.9.0
 pandoc==2.3
 pypandoc==1.11
 tqdm==4.66.1
-sentence_transformers==2.2.2
+sentence_transformers==2.2.2
+numpy>=1.22.2 # not directly required, pinned by Snyk to avoid a vulnerability

examples/langchain-python-rag-websummary/main.py

@@ -1,12 +1,12 @@
-from langchain.llms import Ollama
-from langchain.document_loaders import WebBaseLoader
+from langchain_community.llms import Ollama
+from langchain_community.document_loaders import WebBaseLoader
 from langchain.chains.summarize import load_summarize_chain
 
 loader = WebBaseLoader("https://ollama.com/blog/run-llama2-uncensored-locally")
 docs = loader.load()
 
-llm = Ollama(model="llama2")
+llm = Ollama(model="llama3")
 chain = load_summarize_chain(llm, chain_type="stuff")
 
-result = chain.run(docs)
+result = chain.invoke(docs) 
 print(result)

examples/langchain-python-simple/README.md

@@ -4,10 +4,10 @@ This example is a basic "hello world" of using LangChain with Ollama.
 
 ## Running the Example
 
-1. Ensure you have the `llama2` model installed:
+1. Ensure you have the `llama3` model installed:
 
    ```bash
-   ollama pull llama2
+   ollama pull llama3
    ```
 
 2. Install the Python Requirements.
@@ -21,4 +21,3 @@ This example is a basic "hello world" of using LangChain with Ollama.
    ```bash
    python main.py
    ```
-  

examples/langchain-python-simple/main.py

@@ -1,6 +1,6 @@
 from langchain.llms import Ollama
 
 input = input("What is your question?")
-llm = Ollama(model="llama2")
+llm = Ollama(model="llama3")
 res = llm.predict(input)
 print (res)

examples/modelfile-mario/Modelfile

@@ -1,4 +1,4 @@
-FROM llama2
+FROM llama3
 PARAMETER temperature 1
 SYSTEM """
 You are Mario from super mario bros, acting as an assistant.

examples/modelfile-mario/readme.md

@@ -2,12 +2,12 @@
 
 # Example character: Mario
 
-This example shows how to create a basic character using Llama2 as the base model.
+This example shows how to create a basic character using Llama3 as the base model.
 
 To run this example:
 
 1. Download the Modelfile
-2. `ollama pull llama2` to get the base model used in the model file.
+2. `ollama pull llama3` to get the base model used in the model file.
 3. `ollama create NAME -f ./Modelfile`
 4. `ollama run NAME`
 
@@ -18,7 +18,7 @@ Ask it some questions like "Who are you?" or "Is Peach in trouble again?"
 What the model file looks like:
 
 ```
-FROM llama2
+FROM llama3
 PARAMETER temperature 1
 SYSTEM """
 You are Mario from Super Mario Bros, acting as an assistant.
@@ -28,7 +28,7 @@ You are Mario from Super Mario Bros, acting as an assistant.
 What if you want to change its behaviour?
 
 - Try changing the prompt
-- Try changing the parameters [Docs](https://github.com/jmorganca/ollama/blob/main/docs/modelfile.md)
+- Try changing the parameters [Docs](https://github.com/ollama/ollama/blob/main/docs/modelfile.md)
 - Try changing the model (e.g. An uncensored model by `FROM wizard-vicuna` this is the wizard-vicuna uncensored model )
 
 Once the changes are made,

examples/modelfile-tweetwriter/readme.md

@@ -1,23 +0,0 @@
-# Example Modelfile - Tweetwriter
-
-This simple examples shows what you can do without any code, simply relying on a Modelfile. The file has two instructions:
-
-1. FROM - The From instructions defines the parent model to use for this one. If you choose a model from the library, you can enter just the model name. For all other models, you need to specify the namespace as well. You could also use a local file. Just include the relative path to the converted, quantized model weights file. To learn more about creating that file, see the `import.md` file in the docs folder of this repository.
-2. SYSTEM - This defines the system prompt for the model and overrides the system prompt from the parent model.
-
-## Running the Example
-
-1. Create the model:
-
-   ```bash
-   ollama create tweetwriter
-   ```
-
-2. Enter a topic to generate a tweet about.
-3. Show the Modelfile in the REPL.
-
-   ```bash
-   /show modelfile
-   ```
-
-   Notice that the FROM and SYSTEM match what was in the file. But there is also a TEMPLATE and PARAMETER. These are inherited from the parent model.

examples/python-json-datagenerator/predefinedschema.py

@@ -2,16 +2,16 @@ import requests
 import json
 import random
 
-model = "llama2"
+model = "llama3"
 template = {
-  "firstName": "", 
-  "lastName": "", 
+  "firstName": "",
+  "lastName": "",
   "address": {
-    "street": "", 
-    "city": "", 
-    "state": "", 
+    "street": "",
+    "city": "",
+    "state": "",
     "zipCode": ""
-  }, 
+  },
   "phoneNumber": ""
 }
 

examples/python-json-datagenerator/randomaddresses.py

@@ -12,7 +12,7 @@ countries = [
     "France",
 ]
 country = random.choice(countries)
-model = "llama2"
+model = "llama3"
 
 prompt = f"generate one realistically believable sample data set of a persons first name, last name, address in {country}, and phone number. Do not use common names. Respond using JSON. Key names should have no backslashes, values should use plain ascii with no special characters."
 

examples/python-json-datagenerator/readme.md

@@ -1,15 +1,15 @@
 # JSON Output Example
 
-![llmjson 2023-11-10 15_31_31](https://github.com/jmorganca/ollama/assets/633681/e599d986-9b4a-4118-81a4-4cfe7e22da25)
+![llmjson 2023-11-10 15_31_31](https://github.com/ollama/ollama/assets/633681/e599d986-9b4a-4118-81a4-4cfe7e22da25)
 
 There are two python scripts in this example. `randomaddresses.py` generates random addresses from different countries. `predefinedschema.py` sets a template for the model to fill in.
 
 ## Running the Example
 
-1. Ensure you have the `llama2` model installed:
+1. Ensure you have the `llama3` model installed:
 
    ```bash
-   ollama pull llama2
+   ollama pull llama3
    ```
 
 2. Install the Python Requirements.

examples/python-loganalysis/readme.md

@@ -1,6 +1,6 @@
 # Log Analysis example
 
-![loganalyzer 2023-11-10 08_53_29](https://github.com/jmorganca/ollama/assets/633681/ad30f1fc-321f-4953-8914-e30e24db9921)
+![loganalyzer 2023-11-10 08_53_29](https://github.com/ollama/ollama/assets/633681/ad30f1fc-321f-4953-8914-e30e24db9921)
 
 This example shows one possible way to create a log file analyzer. It uses the model **mattw/loganalyzer** which is based on **codebooga**, a 34b parameter model.
 

examples/python-simplechat/client.py

@@ -2,13 +2,14 @@ import json
 import requests
 
 # NOTE: ollama must be running for this to work, start the ollama app or run `ollama serve`
-model = "llama2"  # TODO: update this for whatever model you wish to use
+model = "llama3"  # TODO: update this for whatever model you wish to use
 
 
 def chat(messages):
     r = requests.post(
         "http://0.0.0.0:11434/api/chat",
         json={"model": model, "messages": messages, "stream": True},
+	stream=True
     )
     r.raise_for_status()
     output = ""