Enable manual test config label selection on Windows

.github/workflows/_win-build.yml

@@ -23,6 +23,18 @@ on:
         description: |
           If this is set, our linter will use this to make sure that every other
           job with the same `sync-tag` is identical.
+      test-matrix:
+        required: false
+        type: string
+        description: |
+          An option JSON description of what test configs to run later on. This
+          is moved here from the Linux test workflow so that we can apply filter
+          logic using test-config labels earlier and skip unnecessary builds
+
+    outputs:
+      test-matrix:
+        value: ${{ inputs.test-matrix }}
+        description: An optional JSON description of what test configs to run later on.
 
 env:
   GIT_DEFAULT_BRANCH: ${{ github.event.repository.default_branch }}
@@ -61,7 +73,17 @@ jobs:
         with:
           github-token: ${{ secrets.GITHUB_TOKEN }}
 
+      # Apply the filter logic to the build step too if the test-config label is already there
+      - name: Select all requested test configurations (if the test matrix is available)
+        id: filter
+        uses: ./.github/actions/filter-test-configs
+        with:
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+          test-matrix: ${{ inputs.test-matrix }}
+
       - name: Build
+        if: steps.filter.outputs.is-test-matrix-empty == 'False' || inputs.test-matrix == ''
+        id: build
         shell: bash
         env:
           PYTORCH_FINAL_PACKAGE_DIR: /c/${{ github.run_id }}/build-results/
@@ -89,6 +111,7 @@ jobs:
 
       # Upload to github so that people can click and download artifacts
       - name: Upload artifacts to s3
+        if: steps.build.outcome != 'skipped'
         uses: seemethere/upload-artifact-s3@v5
         with:
           retention-days: 14
@@ -97,6 +120,7 @@ jobs:
           path: C:\${{ github.run_id }}\build-results
 
       - name: Upload sccache stats
+        if: steps.build.outcome != 'skipped'
         uses: seemethere/upload-artifact-s3@v5
         with:
           s3-prefix: |

.github/workflows/_win-test.yml

@@ -27,11 +27,33 @@ env:
   GIT_DEFAULT_BRANCH: ${{ github.event.repository.default_branch }}
 
 jobs:
+  # This needs to be run right before the test starts so that it can gather the
+  # latest labels from the PR
+  filter:
+    runs-on: [self-hosted, linux.large]
+    outputs:
+      test-matrix: ${{ steps.filter.outputs.test-matrix }}
+      is-test-matrix-empty: ${{ steps.filter.outputs.is-test-matrix-empty }}
+    steps:
+      - name: Checkout PyTorch
+        uses: pytorch/pytorch/.github/actions/checkout-pytorch@master
+        with:
+          fetch-depth: 1
+          submodules: false
+
+      - name: Select all requested test configurations
+        id: filter
+        uses: ./.github/actions/filter-test-configs
+        with:
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+          test-matrix: ${{ inputs.test-matrix }}
+
   test:
-    # Don't run on forked repos.
-    if: github.repository_owner == 'pytorch'
+    needs: filter
+    # Don't run on forked repos or empty test matrix
+    if: github.repository_owner == 'pytorch' && needs.filter.outputs.is-test-matrix-empty == 'False'
     strategy:
-      matrix: ${{ fromJSON(inputs.test-matrix) }}
+      matrix: ${{ fromJSON(needs.filter.outputs.test-matrix) }}
       fail-fast: false
     runs-on: ${{ matrix.runner }}
     timeout-minutes: 300

.github/workflows/periodic.yml

@@ -162,6 +162,12 @@ jobs:
     with:
       build-environment: win-vs2019-cuda11.7-py3
       cuda-version: "11.7"
+      test-matrix: |
+        { include: [
+          { config: "default", shard: 1, num_shards: 2, runner: "windows.8xlarge.nvidia.gpu" },
+          { config: "default", shard: 2, num_shards: 2, runner: "windows.8xlarge.nvidia.gpu" },
+          { config: "force_on_cpu", shard: 1, num_shards: 1, runner: "windows.4xlarge" },
+        ]}
 
   win-vs2019-cuda11_7-py3-test:
     name: win-vs2019-cuda11.7-py3
@@ -170,12 +176,7 @@ jobs:
     with:
       build-environment: win-vs2019-cuda11.7-py3
       cuda-version: "11.7"
-      test-matrix: |
-        { include: [
-          { config: "default", shard: 1, num_shards: 2, runner: "windows.8xlarge.nvidia.gpu" },
-          { config: "default", shard: 2, num_shards: 2, runner: "windows.8xlarge.nvidia.gpu" },
-          { config: "force_on_cpu", shard: 1, num_shards: 1, runner: "windows.4xlarge" },
-        ]}
+      test-matrix: ${{ needs.win-vs2019-cuda11_7-py3-build.outputs.test-matrix }}
 
   ios-12-5-1-x86-64-coreml:
     name: ios-12-5-1-x86-64-coreml

.github/workflows/pull.yml

@@ -237,6 +237,12 @@ jobs:
     with:
       build-environment: win-vs2019-cpu-py3
       cuda-version: cpu
+      test-matrix: |
+        { include: [
+          { config: "default", shard: 1, num_shards: 2, runner: "windows.4xlarge" },
+          { config: "default", shard: 2, num_shards: 2, runner: "windows.4xlarge" },
+          { config: "functorch", shard: 1, num_shards: 1, runner: "windows.4xlarge" },
+        ]}
 
   win-vs2019-cpu-py3-test:
     name: win-vs2019-cpu-py3
@@ -245,12 +251,7 @@ jobs:
     with:
       build-environment: win-vs2019-cpu-py3
       cuda-version: cpu
-      test-matrix: |
-        { include: [
-          { config: "default", shard: 1, num_shards: 2, runner: "windows.4xlarge" },
-          { config: "default", shard: 2, num_shards: 2, runner: "windows.4xlarge" },
-          { config: "functorch", shard: 1, num_shards: 1, runner: "windows.4xlarge" },
-        ]}
+      test-matrix: ${{ needs.win-vs2019-cpu-py3-build.outputs.test-matrix }}
 
   win-vs2019-cuda11_6-py3-build:
     if: github.event_name == 'pull_request'
@@ -260,6 +261,16 @@ jobs:
       build-environment: win-vs2019-cuda11.6-py3
       cuda-version: "11.6"
       sync-tag: win-cuda-build
+      test-matrix: |
+        { include: [
+          { config: "default", shard: 1, num_shards: 5, runner: "windows.8xlarge.nvidia.gpu" },
+          { config: "default", shard: 2, num_shards: 5, runner: "windows.8xlarge.nvidia.gpu" },
+          { config: "default", shard: 3, num_shards: 5, runner: "windows.8xlarge.nvidia.gpu" },
+          { config: "default", shard: 4, num_shards: 5, runner: "windows.8xlarge.nvidia.gpu" },
+          { config: "default", shard: 5, num_shards: 5, runner: "windows.8xlarge.nvidia.gpu" },
+          { config: "functorch", shard: 1, num_shards: 1, runner: "windows.8xlarge.nvidia.gpu" },
+          { config: "force_on_cpu", shard: 1, num_shards: 1, runner: "windows.4xlarge" },
+        ]}
 
   linux-bionic-cuda11_6-py3_10-gcc7-bazel-test:
     name: linux-bionic-cuda11.6-py3.10-gcc7-bazel-test

.github/workflows/trunk.yml

@@ -220,14 +220,6 @@ jobs:
       build-environment: win-vs2019-cuda11.6-py3
       cuda-version: "11.6"
       sync-tag: win-cuda-build
-
-  win-vs2019-cuda11_6-py3-test:
-    name: win-vs2019-cuda11.6-py3
-    uses: ./.github/workflows/_win-test.yml
-    needs: win-vs2019-cuda11_6-py3-build
-    with:
-      build-environment: win-vs2019-cuda11.6-py3
-      cuda-version: "11.6"
       test-matrix: |
         { include: [
           { config: "default", shard: 1, num_shards: 5, runner: "windows.8xlarge.nvidia.gpu" },
@@ -239,6 +231,15 @@ jobs:
           { config: "force_on_cpu", shard: 1, num_shards: 1, runner: "windows.4xlarge" },
         ]}
 
+  win-vs2019-cuda11_6-py3-test:
+    name: win-vs2019-cuda11.6-py3
+    uses: ./.github/workflows/_win-test.yml
+    needs: win-vs2019-cuda11_6-py3-build
+    with:
+      build-environment: win-vs2019-cuda11.6-py3
+      cuda-version: "11.6"
+      test-matrix: ${{ needs.win-vs2019-cuda11_6-py3-build.outputs.test-matrix }}
+
   linux-focal-rocm5_2-py3_7-build:
     name: linux-focal-rocm5.2-py3.7
     uses: ./.github/workflows/_linux-build.yml