Workaround for GitHub MPS environment

It seems that for GitHub Actions MPS-based environment the necessary virtualization technology is not available, so the GPU is visible, but cannot be accessed.
See:
- https://docs.github.com/en/actions/using-github-hosted-runners/using-github-hosted-runners/about-github-hosted-runners#supported-runners-and-hardware-resources
- actions/runner-images#9254
- sktime/pytorch-forecasting#1596
- sktime/pytorch-forecasting#1648
- https://discuss.pytorch.org/t/mps-back-end-out-of-memory-on-github-action/189773/7
- https://github.blog/changelog/2024-01-30-github-actions-introducing-the-new-m1-macos-runner-available-to-open-source/

Author: kyouma

deepxde/backend/pytorch/tensor.py

@@ -21,8 +21,23 @@
     else:
         torch.set_default_tensor_type(torch.cuda.FloatTensor)
 elif torch.backends.mps.is_available():
+    fallback_device = torch.get_default_device()
     torch.set_default_device("mps")
-    torch._dynamo.disable()  # A temporary trick to evade the Pytorch MPS bug (https://github.com/pytorch/pytorch/issues/149184)
+    
+    # As of March 2025, the MacOS X-based GitHub Actions building enviroment sees 
+    # the MPS GPU, but can't access it. Thus as try-except workaround is applied.
+    try:
+        # A temporary trick to evade the Pytorch optimizer bug on MPS
+        # See https://github.com/pytorch/pytorch/issues/149184
+        torch._dynamo.disable()
+
+    except Exception as e:
+        import warnings
+        warnings.warn(
+            f'An MPS GPU has been detected, but cannot be used.'
+            f'Falling back to CPU.\nThe exception message is:\n    {e}'
+        )
+        torch.set_default_device(fallback_device)
 
 
 lib = torch