Test if h265 can work on other GPU runners

Author: scotts

.github/workflows/linux_cuda_wheel.yaml

@@ -56,7 +56,7 @@ jobs:
       build-command: "BUILD_AGAINST_ALL_FFMPEG_FROM_S3=1 ENABLE_CUDA=1 python -m build --wheel -vvv --no-isolation"
 
   install-and-test:
-    runs-on: linux.4xlarge.nvidia.gpu
+    runs-on: linux.g5.4xlarge.nvidia.gpu
     strategy:
       fail-fast: false
       matrix:

test/decoders/test_video_decoder.py

@@ -426,6 +426,7 @@ def test_get_frame_played_at(self, device):
         assert isinstance(decoder.get_frame_played_at(6.02).pts_seconds, float)
         assert isinstance(decoder.get_frame_played_at(6.02).duration_seconds, float)
 
+    @pytest.mark.parametrize("device", cpu_and_cuda())
     def test_get_frame_played_at_h265(self):
         # Non-regression test for https://github.com/pytorch/torchcodec/issues/179
         # We don't parametrize with CUDA because the current GPUs on CI do not