240. MSE Loss (PyTorch)

Implement Mean Squared Error in PyTorch using primitive tensor ops only.

Signature: def mse_loss(y_pred: torch.Tensor, y_true: torch.Tensor) -> torch.Tensor

The rule: you may NOT call F.mse_loss or nn.MSELoss. Implement the formula yourself.

Allowed primitives: **, .mean(), basic arithmetic.

Formula:

mse(y_pred, y_true) = ((y_pred - y_true) ** 2).mean()

Return type — important: unlike the NumPy version which returns a Python float, the PyTorch version should return a 0-dimensional tensor (the result of .mean()). Do not call .item() — the autograd graph would be broken. The harness compares against a scalar value automatically.