Write a custom CUDA kernel to optimize `IpLU`.

Formula:
  f(x) = x                     if x >= 0
  f(x) = x / (1 + |x|^alpha)   if x < 0

Problem Analysis:
1. Memory Bound: This is a point-wise activation. Performance is limited by memory bandwidth.
2. Operator Chaining: The PyTorch implementation using `torch.where` and a chain of `abs`, `pow`, `add`, `div` creates multiple intermediate memory accesses.

Optimization Strategy: Fused Element-wise Kernel with Vectorization

1. One-Thread-per-Element: Map each element to a CUDA thread.

2. Vectorized Loads (float4): Use `float4` to process 128 bits per memory transaction.

3. Fused Branching Logic:
   - For each element `x`, check `if (x < 0)`.
   - If true, compute `denom = 1.0f + __powf(fabsf(x), alpha)`, `result = x / denom`.
   - If false, result is `x`.

4. One-Pass: Fuse all steps into a single read-compute-write kernel.
  
Here's an example to show you the syntax of inline embedding custom CUDA operators in torch: The example given architecture is:   
  
```python
import torch
import torch.nn as nn

BATCH_SIZE = 4096
HIDDEN_DIM = 4096
SHAPE = (BATCH_SIZE, HIDDEN_DIM)

ALPHA_VALUE = 3.0

class IpLU(nn.Module):
    '''
    IpLU (Inverse Polynomial Linear Unit) 
    Formula:
      f(x) = x                     if x >= 0
      f(x) = x / (1 + |x|^alpha)   if x < 0
    '''
    def __init__(self, alpha=2.0):
        super(IpLU, self).__init__()
        self.alpha = alpha

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        pos_part = x
        neg_part = x / (1.0 + torch.abs(x).pow(self.alpha))
        return torch.where(x >= 0, pos_part, neg_part)

class Model(nn.Module):
    def __init__(self, alpha=2.0):
        super(Model, self).__init__()
        self.act = IpLU(alpha=alpha)
    
    def forward(self, x):
        return self.act(x)

def get_inputs():
    input_tensor = torch.randn(SHAPE, dtype=torch.float32) * 2.0
    return [input_tensor.contiguous()]

def get_init_inputs():
    return [ALPHA_VALUE]