Add checks for some workflow jobs (#18583)

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

.github/workflows/self-push.yml

@@ -111,9 +111,24 @@ jobs:
           echo "::set-output name=matrix::$keys"
           echo "::set-output name=test_map::$test_map"
 
+  run_check_runners:
+    name: Check Runners
+    needs: setup
+    strategy:
+      matrix:
+        machine_type: [single-gpu, multi-gpu]
+    runs-on: [self-hosted, docker-gpu, '${{ matrix.machine_type }}']
+    container:
+      image: huggingface/transformers-all-latest-gpu
+      options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/
+    steps:
+      - name: NVIDIA-SMI
+        run: |
+          nvidia-smi
+
   run_tests_single_gpu:
     name: Model tests
-    needs: setup
+    needs: [setup, run_check_runners]
     # `dummy` means there is no test to run
     if: contains(fromJson(needs.setup.outputs.matrix), 'dummy') != true
     strategy:
@@ -198,7 +213,7 @@ jobs:
 
   run_tests_multi_gpu:
     name: Model tests
-    needs: setup
+    needs: [setup, run_check_runners]
     # `dummy` means there is no test to run
     if: contains(fromJson(needs.setup.outputs.matrix), 'dummy') != true
     strategy:
@@ -285,7 +300,7 @@ jobs:
 
   run_tests_torch_cuda_extensions_single_gpu:
     name: Torch CUDA extension tests
-    needs: setup
+    needs: [setup, run_check_runners]
     if: contains(fromJson(needs.setup.outputs.matrix), 'deepspeed') || contains(fromJson(needs.setup.outputs.matrix), 'extended')
     strategy:
       fail-fast: false
@@ -364,7 +379,7 @@ jobs:
 
   run_tests_torch_cuda_extensions_multi_gpu:
     name: Torch CUDA extension tests
-    needs: setup
+    needs: [setup, run_check_runners]
     if: contains(fromJson(needs.setup.outputs.matrix), 'deepspeed') || contains(fromJson(needs.setup.outputs.matrix), 'extended')
     strategy:
       fail-fast: false
@@ -447,12 +462,20 @@ jobs:
     if: always()
     needs: [
         setup,
+        run_check_runners,
         run_tests_single_gpu,
         run_tests_multi_gpu,
         run_tests_torch_cuda_extensions_single_gpu,
         run_tests_torch_cuda_extensions_multi_gpu
     ]
     steps:
+      - name: Preliminary job status
+        shell: bash
+        # For the meaning of these environment variables, see the job `Setup`
+        run: |
+          echo "Setup status: ${{ needs.setup.result }}"
+          echo "Runner status: ${{ needs.run_check_runners.result }}"
+
       # Necessary to get the correct branch name and commit SHA for `workflow_run` event
       # We also take into account the `push` event (we might want to test some changes in a branch)
       - name: Prepare custom environment variables
@@ -498,6 +521,9 @@ jobs:
           CI_TITLE_PUSH: ${{ github.event.head_commit.message }}
           CI_TITLE_WORKFLOW_RUN: ${{ github.event.workflow_run.head_commit.message }}
           CI_SHA: ${{ env.CI_SHA }}
+          SETUP_STATUS: ${{ needs.setup.result }}
+          RUNNER_STATUS: ${{ needs.run_check_runners.result }}
+
         # We pass `needs.setup.outputs.matrix` as the argument. A processing in `notification_service.py` to change
         # `models/bert` to `models_bert` is required, as the artifact names use `_` instead of `/`.
         run: |