Remove only_for=("cpu", "cuda") restriction from
TestNNParametrizationDevice to enable tests on all accelerator devices.
test_weight_norm_parametrization uses only device-agnostic operations:
torch.randn(..., device=device)
nn.Linear(..., device=device)
torch.nn.utils.parametrizations.weight_norm()
This allows the test to run on any registered...
🛡️ VERIFIED CYBER INTELLIGENCE ID: #3526013