use instance with more gpu memory and only run py38 and cu116 on PR CI

Author: osalpekar

.github/workflows/test-linux-gpu.yml

@@ -16,12 +16,12 @@ jobs:
   tests:
     strategy:
       matrix:
-        python_version: ["3.7", "3.8", "3.9", "3.10"]
-        cuda_arch_version: ["11.6", "11.7"]
+        python_version: ["3.8"]
+        cuda_arch_version: ["11.6"]
       fail-fast: false
     uses: pytorch/test-infra/.github/workflows/linux_job.yml@main
     with:
-      runner: linux.8xlarge.nvidia.gpu
+      runner: linux.g5.4xlarge.nvidia.gpu
       repository: pytorch/vision
       gpu-arch-type: cuda
       gpu-arch-version: ${{ matrix.cuda_arch_version }}