🐛[BUG]: MeshGraphNet tests fail if default norm is TELayerNorm

[coreyjadams]

Version

0.10.0

On which installation method(s) does this occur?

No response

Describe the issue

MeshGraphNet is approximately 2x faster in some scenarios using TELayerNorm instead of LayerNorm. TELayerNorm is not available for CPU.

In the pending PR #787 , MeshGraphNet will catch and replace with LayerNorm if cuda is not available. But for the tests, cuda is available on the CPU tests, just not used.

To fix this issue, we need to update the tests to use LayerNorm on cpu devices.

Minimum reproducible example

Relevant log output

Environment details

[coreyjadams]

FYI @mnabian @ktangsali I plan to make these changes after the RC candidate.