Migrate the CI runners to the new clusters (#33849)

* try fixing push-ci

* move to new runners

* move benchmark.yml to new runners

* move doctest_job.yml to new runners

* move doctests.yml to new runners

* move push-important-models.yml to new runners

* move self-pr-slow-ci.yml to new runners

* fix typo

Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com>

* fix working directory

Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com>

* fix working directory

Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com>

* improve code

Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com>

---------

Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com>

.github/workflows/doctest_job.yml

@@ -27,7 +27,8 @@ jobs:
       fail-fast: false
       matrix:
         split_keys: ${{ fromJson(inputs.split_keys) }}
-    runs-on: [single-gpu, nvidia-gpu, t4, ci]
+    runs-on: 
+      group: aws-g4dn-2xlarge-cache
     container:
       image: huggingface/transformers-all-latest-gpu
       options: --gpus 0 --shm-size "16gb" --ipc host -v /mnt/cache/.cache/huggingface:/mnt/cache/