Resolve #2431: Remove macos-13 from tests (#2906)

* Resolve #2431: Remove macos-13 from tests

As discussed in #2431 since there are no better macos runners available
and the intel macos runners are failing for quite some time now we now
have little choice other than to stop using macos in the CI.

Author: githubnemo

.github/workflows/tests.yml

@@ -41,10 +41,7 @@ jobs:
       fail-fast: false
       matrix:
         python-version: ["3.10", "3.11", "3.12", "3.13"]
-        os: ["ubuntu-latest", "macos-13", "windows-latest"]
-        exclude:
-          - os: macos-13
-            python-version: "3.13"
+        os: ["ubuntu-latest", "windows-latest"]
     runs-on: ${{ matrix.os }}
     steps:
       - uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8  # v5.0.0
@@ -82,30 +79,12 @@ jobs:
           # cpu version of pytorch
           pip install -e .[test]
       - name: Test with pytest
-        # MacOS tests are currently too flaky and will fail almost each time. Thus, continue (green checkmark) even if
-        # they fail, but add a notice so that the failure is not completely silent
-        continue-on-error: ${{ matrix.os == 'macos-13' }}
         shell: bash
         env:
           HF_TOKEN: ${{ secrets.HF_TOKEN }}
           TRANSFORMERS_IS_CI: 1
         run: |
-          set +e
           make test
-          status=$?
-          # Post a notice only if this is macOS AND tests failed
-          if [ "$status" -ne 0 ] && [ "${{ matrix.os }}" = "macos-13" ]; then
-            {
-              echo "## ⚠️ macOS tests failed"
-              echo ""
-              echo "- OS: ${{ matrix.os }}"
-              echo "- Python: ${{ matrix.python-version }}"
-              echo ""
-              echo "Check the logs from this step for details."
-            } >> "$GITHUB_STEP_SUMMARY"
-          fi
-          # Return the real status. On macOS this won't fail the job because of continue-on-error.
-          exit $status
       - name: Dump cache content and diff
         # This is just debug info so that we can monitor if the model cache diverges substantially
         # over time and what the diverging model is.