ci: update caching to resolve concurrency (#2869)

* ci: update caching to resolve concurrency
* Apply suggestions from code review

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

.github/actions/pull-caches/action.yml

@@ -18,6 +18,14 @@ inputs:
     description: cache restore/dump key
     required: false
     default: "pypi-packages"
+  cache-torch-HF:
+    description: "cache torch and HF"
+    required: false
+    default: "true"
+  cache-references:
+    description: "cache metrics references"
+    required: false
+    default: "false"
 
 runs:
   using: "composite"
@@ -67,6 +75,7 @@ runs:
       shell: bash
 
     - name: Cache Torch & HF
+      if: inputs.cache-torch-HF == 'true' # since the input is string
       continue-on-error: true
       uses: actions/cache/restore@v3
       with:
@@ -75,6 +84,7 @@ runs:
         key: ci-caches
 
     - name: Restored Torch & HF
+      if: inputs.cache-torch-HF == 'true' # since the input is string
       run: |
         mkdir -p $CACHES_DIR
         pip install -q py-tree
@@ -83,14 +93,15 @@ runs:
 
     - name: Cache References
       # do not use this cache for dispatch and crone, to enable rebuild caches if needed
-      if: github.event_name != 'workflow_dispatch' && github.event_name != 'schedule'
+      if: github.event_name != 'workflow_dispatch' && github.event_name != 'schedule' && inputs.cache-references == 'true'
       continue-on-error: true
       uses: actions/cache/restore@v3
       with:
         path: tests/_cache-references
         key: cache-references
 
     - name: Restored References
+      if: inputs.cache-references == 'true' # since the input is string
       continue-on-error: true
       working-directory: tests/
       run: |

.github/actions/push-caches/action.yml

@@ -14,6 +14,18 @@ inputs:
     description: location to pull PyTorch from
     required: false
     default: "https://download.pytorch.org/whl/cpu/torch_stable.html"
+  cache-artifact-appendix:
+    description: "unique name or running index"
+    required: false
+    default: ""
+  cache-torch-HF:
+    description: "cache torch and HF"
+    required: false
+    default: "true"
+  cache-references:
+    description: "cache metrics references"
+    required: false
+    default: "false"
 
 runs:
   using: "composite"
@@ -23,66 +35,50 @@ runs:
       shell: bash
 
     - name: Freeze local emv.
+      if: inputs.cache-artifact-appendix != ''
       run: |
         pip freeze > requirements.dump
         cat requirements.dump
       shell: bash
 
-    #- name: Filter self pkg
-    #  run: |
-    #    import os
-    #    fp = 'requirements.dump'
-    #    with open(fp) as fopen:
-    #      lines = [ln.strip() for ln in fopen.readlines()]
-    #    lines = [ln.split('+')[0] for ln in lines if '-e ' not in ln]
-    #    with open(fp, 'w') as fwrite:
-    #      fwrite.writelines([ln + os.linesep for ln in lines])
-    #  shell: python
-
     - name: Dump wheels
+      if: inputs.cache-artifact-appendix != ''
       run: |
         pip wheel -r requirements/_devel.txt --prefer-binary \
-          --wheel-dir=.pip-wheels \
+          --wheel-dir=_pip-wheels \
           -f ${{ inputs.torch-url }} -f ${{ inputs.pypi-dir }}
-        ls -lh .pip-wheels
+        ls -lh _pip-wheels
       shell: bash
 
-    - name: Cache pull packages
-      uses: actions/cache/restore@v3
-      with:
-        enableCrossOsArchive: true
-        path: ${{ inputs.pypi-dir }}
-        key: ${{ inputs.pypi-key }}
-
-    - name: Find diff
-      id: wheels-diff
+    - name: Move new packages to staging
+      if: inputs.cache-artifact-appendix != ''
       run: |
-        import os, glob
-        wheels = [os.path.basename(p) for p in glob.glob(".pip-wheels/*")]
-        pkgs = [os.path.basename(p) for p in glob.glob("${{ inputs.pypi-dir }}/*")]
-        diff = [w for w in wheels if w not in pkgs]
-        print(diff)
-        with open(os.environ['GITHUB_OUTPUT'], 'a') as fh:
-            print(f'count-new={len(diff)}', file=fh)
-      shell: python
-
-    - run: cp .pip-wheels/* ${{ inputs.pypi-dir }}
-      if: ${{ steps.wheels-diff.outputs.count-new != 0 }}
+        mkdir -p _pip-staging
+        python .github/assistant.py move_new_packages \
+          --dir-cache="${{ inputs.pypi-dir }}" \
+          --dir_local="_pip-wheels" \
+          --dir_staging="_pip-staging"
+        ls -lh _pip-staging/
+        # count files in the staging dir
+        file_count=$(ls -1 "_pip-staging/" | wc -l)
+        echo "NUM_PACKAGES=$file_count" >> $GITHUB_ENV
       shell: bash
 
-    - name: Cache push packages
-      if: ${{ steps.wheels-diff.outputs.count-new != 0 }}
-      uses: actions/cache/save@v3
+    - name: Upload new packages
+      if: inputs.cache-artifact-appendix != '' && env.NUM_PACKAGES != 0
+      uses: actions/upload-artifact@v4
       with:
-        enableCrossOsArchive: true
-        path: ${{ inputs.pypi-dir }}
-        key: ${{ inputs.pypi-key }}
+        name: ${{ inputs.pypi-key }}-run-${{ inputs.cache-artifact-appendix }}
+        path: _pip-staging
+        retention-days: 1
 
     - name: Post Torch & HF
+      if: inputs.cache-torch-HF == 'true' # since the input is string
       run: py-tree $CACHES_DIR
       shell: bash
 
     - name: Cache Torch & HF
+      if: inputs.cache-torch-HF == 'true' # since the input is string
       continue-on-error: true
       uses: actions/cache/save@v3
       with:
@@ -91,13 +87,16 @@ runs:
         key: ci-caches
 
     - name: Cache references
+      if: inputs.cache-references == 'true' # since the input is string
       continue-on-error: true
       uses: actions/cache/save@v3
       with:
         #enableCrossOsArchive: true
         path: tests/_cache-references
         key: cache-references
 
-    - name: Post References
-      run: py-tree tests/_cache-references/ --show_hidden
-      shell: bash
+    #- name: Post References
+    #  # This print taken soo many lines, so it is commented out
+    #  if: inputs.cache-references == 'true' # since the input is string
+    #  run: py-tree tests/_cache-references/ --show_hidden
+    #  shell: bash

.github/assistant.py

@@ -179,6 +179,23 @@ def _crop_path(fname: str, paths: list[str]) -> str:
             raise ValueError(f"Missing following paths: {not_exists}")
         return " ".join(test_modules)
 
+    @staticmethod
+    def move_new_packages(dir_cache: str, dir_local: str, dir_staging: str) -> None:
+        """Move unique packages from local folder to staging."""
+        assert os.path.isdir(dir_cache), f"Missing folder with saved packages: '{dir_cache}'"  # noqa: S101
+        assert os.path.isdir(dir_local), f"Missing folder with local packages: '{dir_local}'"  # noqa: S101
+        assert os.path.isdir(dir_staging), f"Missing folder for staging: '{dir_staging}'"  # noqa: S101
+
+        import shutil
+
+        for pkg in os.listdir(dir_local):
+            if not os.path.isfile(pkg):
+                continue
+            if pkg in os.listdir(dir_cache):
+                continue
+            logging.info(f"Moving '{pkg}' to staging...")
+            shutil.move(os.path.join(dir_cache, pkg), os.path.join(dir_staging, pkg))
+
 
 if __name__ == "__main__":
     logging.basicConfig(level=logging.INFO)

.github/workflows/_merge_cache.yml

@@ -0,0 +1,57 @@
+name: Collect new packages and upload cache
+
+on:
+  workflow_call:
+    inputs:
+      pypi-key:
+        description: cache restore/dump key
+        required: false
+        type: string
+        default: "pypi-packages"
+      pypi-dir:
+        description: location of local PyPI cache
+        required: false
+        type: string
+        default: "_ci-cache_PyPI"
+      cache-artifact-appendix:
+        description: "unique name for the job"
+        required: true
+        type: string
+
+jobs:
+  merge-caches:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Download 📥 artifacts
+        uses: actions/download-artifact@v4
+        with:
+          pattern: ${{ inputs.pypi-key }}-run-${{ inputs.cache-artifact-appendix }}*
+          merge-multiple: true
+          path: _local-packages
+      - name: Cache pull packages
+        uses: actions/cache/restore@v3
+        with:
+          enableCrossOsArchive: true
+          path: ${{ inputs.pypi-dir }}
+          key: ${{ inputs.pypi-key }}
+
+      - name: show 📦
+        run: |
+          # create the directory if it doesn't exist - no artifact were found
+          mkdir -p _local-packages
+          ls -lh _local-packages
+          ls -lh ${{ inputs.pypi-dir }}
+          # count files in the staging dir
+          file_count=$(ls -1 "_local-packages/" | wc -l)
+          echo "NUM_PACKAGES=$file_count" >> $GITHUB_ENV
+      - name: Move collected 📦
+        if: env.NUM_PACKAGES != 0
+        run: mv _local-packages/* ${{ inputs.pypi-dir }}
+
+      - name: Cache push packages
+        if: env.NUM_PACKAGES != 0
+        uses: actions/cache/save@v3
+        with:
+          enableCrossOsArchive: true
+          path: ${{ inputs.pypi-dir }}
+          key: ${{ inputs.pypi-key }}

.github/workflows/ci-tests.yml

@@ -20,6 +20,9 @@ defaults:
   run:
     shell: bash
 
+env:
+  PYPI_CACHE_DIR: "_ci-cache_PyPI"
+
 jobs:
   check-diff:
     if: github.event.pull_request.draft == false
@@ -59,10 +62,10 @@ jobs:
           - { os: "windows-2022", python-version: "3.11", pytorch-version: "2.6.0" }
     env:
       FREEZE_REQUIREMENTS: ${{ ! (github.ref == 'refs/heads/master' || startsWith(github.ref, 'refs/heads/release/')) }}
-      PYPI_CACHE_DIR: "_ci-cache_PyPI"
       TOKENIZERS_PARALLELISM: false
       TEST_DIRS: ${{ needs.check-diff.outputs.test-dirs }}
-      PIP_EXTRA_INDEX_URL: "--find-links https://download.pytorch.org/whl/cpu/torch_stable.html"
+      PIP_EXTRA_INDEX_URL: "--find-links=https://download.pytorch.org/whl/cpu/torch_stable.html"
+      UNITTEST_TIMEOUT: "" # by default, it is not set
 
     # Timeout: https://stackoverflow.com/a/59076067/4521646
     # seems that macOS jobs take much more than orger OS
@@ -98,6 +101,7 @@ jobs:
           requires: ${{ matrix.requires }}
           pytorch-version: ${{ matrix.pytorch-version }}
           pypi-dir: ${{ env.PYPI_CACHE_DIR }}
+          cache-references: true
 
       - name: Switch to PT test URL
         if: ${{ matrix.pytorch-version == '2.6.0' }}
@@ -107,7 +111,7 @@ jobs:
         run: |
           pip --version
           pip install -e . -U "setuptools==69.5.1" -r requirements/_doctest.txt \
-            $PIP_EXTRA_INDEX_URL --find-links $PYPI_CACHE_DIR
+            $PIP_EXTRA_INDEX_URL --find-links="$PYPI_CACHE_DIR"
           pip list
 
       - name: DocTests
@@ -126,7 +130,7 @@ jobs:
               python adjust-torch-versions.py $fpath
           done
           pip install --requirement requirements/_devel.txt -U \
-            $PIP_EXTRA_INDEX_URL --find-links $PYPI_CACHE_DIR
+            $PIP_EXTRA_INDEX_URL --find-links="$PYPI_CACHE_DIR"
           pip list
 
       - name: set special vars for PR
@@ -201,7 +205,7 @@ jobs:
         uses: codecov/codecov-action@v5
         with:
           token: ${{ secrets.CODECOV_TOKEN }}
-          file: tests/coverage.xml
+          files: "tests/coverage.xml"
           flags: cpu,${{ runner.os }},python${{ matrix.python-version }},torch${{ steps.info.outputs.TORCH }}
           env_vars: OS,PYTHON
           name: codecov-umbrella
@@ -213,6 +217,8 @@ jobs:
         uses: ./.github/actions/push-caches
         with:
           pypi-dir: ${{ env.PYPI_CACHE_DIR }}
+          cache-artifact-appendix: ${{ github.run_id }}-${{ strategy.job-index }}
+          cache-references: true
 
   testing-guardian:
     runs-on: ubuntu-latest
@@ -227,3 +233,11 @@ jobs:
         if: contains(fromJSON('["cancelled", "skipped"]'), needs.pytester.result)
         timeout-minutes: 1
         run: sleep 90
+
+  merge-pkg-artifacts:
+    needs: pytester
+    if: success()
+    uses: ./.github/workflows/_merge_cache.yml
+    with:
+      pypi-dir: "_ci-cache_PyPI"
+      cache-artifact-appendix: ${{ github.run_id }}-${{ strategy.job-index }}

tests/unittests/_helpers/wrappers.py

@@ -10,7 +10,6 @@
     "We couldn't connect to",
     "Connection error",
     "Can't load",
-    "`nltk` resource `punkt` is",
 )
 
 

tests/unittests/utilities/test_auc.py → tests/unittests/classification/test_auc.py

@@ -19,6 +19,7 @@
 from sklearn.metrics import auc as _sk_auc
 from torch import Tensor, tensor
 from torchmetrics.utilities.compute import auc
+
 from unittests import NUM_BATCHES
 from unittests._helpers import seed_all
 from unittests._helpers.testers import MetricTester