Docker Release Images v2 (#15616)

### Ticket #12496 #12495 (describes agreed upon naming scheme) ### Problem description Allow users to get started with TT-Metal stack within two command lines. ``` docker pull ghcr.io/tenstorrent/tt-metal/tt-metalium/ubuntu-20.04-amd64-release/wormhole_b0:latest docker run --rm -it -t ghcr.io/tenstorrent/tt-metal/tt-metalium/ubuntu-20.04-amd64-release/wormhole_b0:latest python3 -c "import ttnn" ``` ### What's changed Add a job for the package release to also build/push a release image. ### Checklist - [x] Post commit CI passes https://github.com/tenstorrent/tt-metal/actions/runs/12127666954 - [ ] Blackhole Post commit (if applicable) - [ ] Model regression CI testing passes (if applicable) - [ ] Device performance regression CI testing passes (if applicable) - [ ] New/Existing tests provide coverage for changes
tenstorrent · Dec 6, 2024 · dc6d684 · dc6d684
1 parent 317d346
commit dc6d684
Show file tree

Hide file tree

Showing 12 changed files with 199 additions and 41 deletions.
diff --git a/.github/actions/docker-run/action.yml b/.github/actions/docker-run/action.yml
@@ -8,7 +8,7 @@ inputs:
   docker_os_arch:
     description: 'Docker image architecture'
     required: false
-    default: ubuntu-20.04-amd64
+    default: tt-metalium/ubuntu-20.04-amd64
   docker_username:
     description: docker login username
     required: true
@@ -89,7 +89,7 @@ runs:
           set -eu
 
           install_wheel=${{ inputs.install_wheel }}
-          if [[ "${install_wheel,,}" == "true" ]]; then
+          if [ "${install_wheel,,}" == "true" ]; then
             WHEEL_FILENAME=$(ls -1 *.whl)
             pip3 install "$WHEEL_FILENAME"
           fi

diff --git a/.github/actions/generate-docker-tag/action.yml b/.github/actions/generate-docker-tag/action.yml
@@ -24,8 +24,8 @@ runs:
     - name: Determine Full Docker Image Tag
       shell: bash
       run: |
-        echo "TT_METAL_DOCKER_IMAGE_TAG=ghcr.io/${{ github.repository }}/tt-metalium/${{ inputs.image }}:${{ env.IMAGE_TAG }}" >> $GITHUB_ENV
-        echo "TT_METAL_REF_IMAGE_TAG=ghcr.io/${{ github.repository }}/tt-metalium/${{ inputs.image }}:latest" >> $GITHUB_ENV
+        echo "TT_METAL_DOCKER_IMAGE_TAG=ghcr.io/${{ github.repository }}/${{ inputs.image }}:${{ env.IMAGE_TAG }}" >> $GITHUB_ENV
+        echo "TT_METAL_REF_IMAGE_TAG=ghcr.io/${{ github.repository }}/${{ inputs.image }}:latest" >> $GITHUB_ENV
     - name: Output Docker Image Tag
       shell: bash
       run: |

diff --git a/.github/workflows/build-artifact.yaml b/.github/workflows/build-artifact.yaml
@@ -99,7 +99,7 @@ jobs:
         id: generate-docker-tag
         uses: ./.github/actions/generate-docker-tag
         with:
-          image: ${{ inputs.os }}
+          image: tt-metalium/${{ inputs.os }}
       - name: Docker login
         uses: docker/login-action@v3
         with:

diff --git a/.github/workflows/build-docker-artifact.yaml b/.github/workflows/build-docker-artifact.yaml
@@ -53,7 +53,7 @@ jobs:
       - name: Determine docker image tag
         uses: ./.github/actions/generate-docker-tag
         with:
-          image: ${{ inputs.os }}
+          image: tt-metalium/${{ inputs.os }}
       - name: Build Docker image and push to GHCR
         if: steps.changed-files-specific.outputs.any_changed == 'true'
         uses: docker/build-push-action@v6

diff --git a/.github/workflows/build.yaml b/.github/workflows/build.yaml
@@ -50,14 +50,13 @@ jobs:
         with:
           docker_username: ${{ github.actor }}
           docker_password: ${{ secrets.GITHUB_TOKEN }}
-          docker_image_arch: ${{ inputs.arch }}
           docker_opts: |
             -e ARCH_NAME=${{ matrix.arch }}
             --group-add 1457
             -v /home/ubuntu/.ccache-ci:/home/ubuntu/.ccache
             -e CCACHE_DIR=/home/ubuntu/.ccache
             -v /mnt/MLPerf/ccache:/mnt/MLPerf/ccache
-          docker_os_arch: ${{ matrix.build.os }}-amd64
+          docker_os_arch: tt-metalium/${{ matrix.build.os }}-amd64
           run_args: |
             set -eu # basic shell hygiene
             set -x

diff --git a/.github/workflows/code-analysis.yaml b/.github/workflows/code-analysis.yaml
@@ -54,7 +54,7 @@ jobs:
         id: generate-docker-tag
         uses: ./.github/actions/generate-docker-tag
         with:
-          image: ${{ inputs.os }}
+          image: tt-metalium/${{ inputs.os }}
       - name: Docker login
         uses: docker/login-action@v3
         with:

diff --git a/.github/workflows/cpp-ttnn-project.yaml b/.github/workflows/cpp-ttnn-project.yaml
@@ -27,7 +27,7 @@ jobs:
         id: generate-docker-tag
         uses: ./.github/actions/generate-docker-tag
         with:
-          image: ubuntu-22.04-amd64
+          image: tt-metalium/ubuntu-22.04-amd64
       - name: Docker login
         uses: docker/login-action@v3
         with:

diff --git a/.github/workflows/package-and-release.yaml b/.github/workflows/package-and-release.yaml
@@ -180,32 +180,16 @@ jobs:
             infra/machine_setup/scripts/setup_hugepages.py
             metal_libs-*+*.whl
           fail_on_unmatched_files: true
-  create-docker-image:
+  create-docker-release-image:
     needs: [
       create-tag,
       create-and-upload-draft-release
     ]
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v4
-      - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3
-      - name: Login to GitHub Container Registry
-        uses: docker/login-action@v3
-        with:
-          registry: https://ghcr.io
-          username: ${{ github.actor }}
-          password: ${{ secrets.GITHUB_TOKEN }}
-      - name: Build and push
-        env:
-          TT_METAL_DOCKER_IMAGE: tt-metalium/ubuntu-20.04-amd64
-        uses: docker/build-push-action@v6
-        with:
-          push: true
-          tags: ghcr.io/${{ github.repository }}/tt-metalium/ubuntu-20.04-amd64:${{ needs.create-tag.outputs.version }}-dev
-          context: .
-          file: dockerfile/ubuntu-20.04-amd64.Dockerfile
+    uses: ./.github/workflows/publish-release-image.yaml
+    secrets: inherit
+    with:
+      version: ${{ needs.create-tag.outputs.version }}
+      is_major_version:  ${{ needs.get-params.outputs.is-release-candidate !='true' && needs.get-params.outputs.should-create-release == 'true' }}
   release-docs:
     needs: [
       get-params,

diff --git a/.github/workflows/publish-release-image-wrapper.yaml b/.github/workflows/publish-release-image-wrapper.yaml
@@ -1,12 +1,28 @@
 name: "Create and Publish Release Docker Image"
 
 on:
-  workflow_call:
   workflow_dispatch:
-
 jobs:
-  to_be_filled_out:
-    steps:
-      - name: This workflow will be filled out in https://github.com/tenstorrent/tt-metal/pull/15013
-        run: |
-          echo "NOOP"
+  build-artifact:
+    uses: ./.github/workflows/build-artifact.yaml
+    secrets: inherit
+  build-wheels:
+    needs: build-artifact
+    strategy:
+      matrix:
+        # Since pre-compiled builds only run on 20.04, we can only test on 20.04 for now
+        # The full 22.04 flow can be tested without precompiled
+        os: [ubuntu-20.04]
+        arch: [grayskull, wormhole_b0]
+    uses: ./.github/workflows/_build-wheels-impl.yaml
+    with:
+      os: ${{ matrix.os }}
+      arch: ${{ matrix.arch }}
+      from-precompiled: true
+  publish-release-image:
+    needs: build-wheels
+    uses: ./.github/workflows/publish-release-image.yaml
+    secrets: inherit
+    with:
+      version: dev-${GITHUB_REF_NAME//\//-}
+      is_major_version: false
diff --git a/.github/workflows/publish-release-image.yaml b/.github/workflows/publish-release-image.yaml
@@ -0,0 +1,131 @@
+name: "[internal] Create and Publish Release Docker Image"
+
+on:
+  workflow_call:
+    inputs:
+      version:
+        required: true
+        type: string
+      is_major_version:
+        required: true
+        type: boolean
+        default: false
+      timeout:
+        required: false
+        type: number
+        default: 10
+jobs:
+  create-docker-release-image:
+    strategy:
+      matrix:
+        os: [ubuntu-20.04]
+        arch: [grayskull, wormhole_b0]
+    runs-on:
+      - build-docker
+      - in-service
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+      - name: Login to GitHub Container Registry
+        uses: docker/login-action@v3
+        with:
+          registry: https://ghcr.io
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+      - name: Download wheels
+        uses: actions/download-artifact@v4
+        with:
+          name: eager-dist-${{ matrix.os }}-${{ matrix.arch }}
+      - name: Get the name of the wheel and set up env variables
+        id: generate-tag-name
+        run: |
+          echo "WHEEL_FILENAME=$(ls -1 *.whl)" >> $GITHUB_ENV
+          REPO_IMAGE_NAME=ghcr.io/${{ github.repository }}/tt-metalium-${{ matrix.os }}-amd64-release/${{ matrix.arch }}
+          echo "REPO_IMAGE_NAME=$REPO_IMAGE_NAME" >> $GITHUB_ENV
+          TAG_NAME=$REPO_IMAGE_NAME:${{ inputs.version }}
+          echo "TAG_NAME=$TAG_NAME" >> $GITHUB_ENV
+      - name: Build and push
+        uses: docker/build-push-action@v6
+        with:
+          push: true
+          build-args: |
+            WHEEL_FILENAME=${{ env.WHEEL_FILENAME }}
+            BASE_IMAGE_NAME=tt-metalium/${{ matrix.os }}-amd64
+          tags: ${{ env.TAG_NAME }}
+          context: .
+          file: dockerfile/release.Dockerfile
+  smoke-test-docker-image:
+    needs: create-docker-release-image
+    strategy:
+      matrix:
+        os: [ubuntu-20.04]
+        test_group:
+          [
+              {
+                arch: grayskull,
+                runs-on: ["cloud-virtual-machine", "E150", "in-service"],
+                cmd: pytest tests/end_to_end_tests,
+              },
+              {
+                arch: wormhole_b0,
+                runs-on: ["cloud-virtual-machine", "N150", "in-service"],
+                cmd: pytest tests/end_to_end_tests,
+              },
+              {
+                arch: wormhole_b0,
+                runs-on: ["cloud-virtual-machine", "N300", "in-service"],
+                cmd: pytest tests/end_to_end_tests,
+              },
+          ]
+    env:
+      ARCH_NAME: ${{ matrix.test_group.arch }}
+      LOGURU_LEVEL: INFO
+    runs-on: ${{ matrix.test_group.runs-on }}
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Run smoke test on the image
+        timeout-minutes: ${{ inputs.timeout }}
+        uses: ./.github/actions/docker-run
+        with:
+          docker_os_arch: tt-metalium-${{ matrix.os }}-amd64-release/${{ matrix.test_group.arch }}
+          docker_password: ${{ secrets.GITHUB_TOKEN }}
+          run_args: |
+            ${{ matrix.test_group.cmd }}
+  tag-docker-image-as-latest:
+    needs: [smoke-test-docker-image, create-docker-release-image]
+    strategy:
+      matrix:
+        os: [ubuntu-20.04]
+        arch: [grayskull, wormhole_b0]
+    runs-on:
+      - build-docker
+      - in-service
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+      - name: Docker login
+        uses: docker/login-action@v3
+        with:
+          registry: https://ghcr.io
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+      - name: Tag latest if this is a major version release
+        run: |
+          set -eu # basic shell hygiene
+          LATEST_TAG=latest
+          if [ "${{  inputs.is_major_version }}" = "true" ]; then
+            LATEST_TAG=latest
+          else
+            LATEST_TAG=latest-rc
+          fi
+          echo "Determined that the current tag is " $LATEST_TAG
+          REPO_IMAGE_NAME=ghcr.io/${{ github.repository }}/tt-metalium-${{ matrix.os }}-amd64-release/${{ matrix.arch }}
+          TAG_NAME=$REPO_IMAGE_NAME:${{ inputs.version }}
+          docker pull $TAG_NAME
+          echo "Tagging the image as " $REPO_IMAGE_NAME:$LATEST_TAG
+          docker tag $TAG_NAME $REPO_IMAGE_NAME:$LATEST_TAG
+          echo "Pushing image with tag " $REPO_IMAGE_NAME:$LATEST_TAG
+          docker push $REPO_IMAGE_NAME:$LATEST_TAG
diff --git a/INSTALLING.md b/INSTALLING.md
@@ -66,14 +66,14 @@ sudo -E python3 setup_hugepages.py enable && sudo -E python3 setup_hugepages.py
 
 > [!NOTE]
 >
-> You may choose to install from either source or a Python wheel.
+> You may choose to install from either source, a Python wheel, or Docker release image.
 >
 > However, no matter your method, in order to use our pre-built models or to
 > follow along with the documentation and tutorials to get started, you will
 > still need the source code.
 >
 > If you do not want to use the models or follow the tutorials and want to
-> immediately start using the API, you may install just the wheel.
+> immediately start using the API, you may install just the wheel or get the release Docker container.
 
 1. Install git and git-lfs.
 
@@ -157,6 +157,24 @@ sudo apt-get install cpufrequtils
 sudo cpupower frequency-set -g performance
 ```
 
+#### Option 3: From Docker Release Image
+
+Download the latest Docker release from our [Docker registry](https://github.com/orgs/tenstorrent/packages?q=tt-metalium-ubuntu&tab=packages&q=tt-metalium-ubuntu-20.04-amd64-release) page for
+the particular Tenstorrent card architecture that you have installed on your
+system. (ie. Grayskull, Wormhole, etc)
+
+```sh
+docker pull ghcr.io/tenstorrent/tt-metal/tt-metalium-ubuntu-20.04-amd64-release/<arch_name>:latest-rc
+docker run --it --rm -v /dev/hugepages-1G:/dev/hugepages-1G --device /dev/tenstorrent ghcr.io/tenstorrent/tt-metal/tt-metalium-ubuntu-20.04-amd64-release/<arch_name>:latest-rc bash
+```
+where `arch_name` is one of `grayskull`, `wormhole_b0`, or `blackhole`,
+depending on your Tenstorrent card type.
+
+When inside of the container,
+```sh
+python3 -c "import ttnn"
+```
+
 2. Start coding
 
 To verify your installation, try the executing an example:

diff --git a/dockerfile/release.Dockerfile b/dockerfile/release.Dockerfile
@@ -0,0 +1,10 @@
+ARG BASE_IMAGE_NAME=tt-metalium/ubuntu-20.04-amd64
+#
+# Currently the release image uses the base image which is also the build image.
+# However, in the future, we could point a true base image that is a base for both releases and builds.
+# This work is described in https://github.com/tenstorrent/tt-metal/issues/11974
+FROM ghcr.io/tenstorrent/tt-metal/$BASE_IMAGE_NAME
+
+ARG WHEEL_FILENAME
+ADD $WHEEL_FILENAME $WHEEL_FILENAME
+RUN pip3 install $WHEEL_FILENAME