238. ELU and SELU (PyTorch)

Implement ELU and SELU in PyTorch using primitive tensor ops only.

Signature: def elu_selu(x: torch.Tensor, mode: str = 'elu', alpha: float = 1.0) -> torch.Tensor

The rule: you may NOT call F.elu, F.selu, nn.ELU, or nn.SELU. We verify your output matches F.elu(x, alpha) and F.selu(x) respectively.

Allowed primitives: .exp(), torch.where, basic arithmetic.

Formulas:

ELU(x)  = x                              if x > 0
        = alpha * (exp(x) - 1)           if x <= 0

SELU(x) = scale * x                      if x > 0
        = scale * alpha * (exp(x) - 1)   if x <= 0

scale = 1.0507009873554804
alpha = 1.6732631921096593   (SELU's fixed alpha; ignore the function arg in 'selu' mode)

PyTorch idioms vs the NumPy version:

• torch.where(cond, a, b) selects element-wise. The condition must be a boolean tensor — x > 0 produces one automatically.
• Both branches are evaluated (PyTorch does not short-circuit in torch.where). For x.exp() on positive inputs you'll compute values that are then discarded — this is fine, just be aware that gradients flow through both branches.