fix: rewrite flux swiglu activation to avoid gather op in neuron IR

Author: JingyaHuang

src/diffusers/models/transformers/transformer_flux2.py

@@ -291,8 +291,8 @@ def __init__(self):
         self.gate_fn = nn.SiLU()
 
     def forward(self, x: torch.Tensor) -> torch.Tensor:
-        x1, x2 = x.chunk(2, dim=-1)
-        x = self.gate_fn(x1) * x2
+        half = x.shape[-1] // 2
+        x = self.gate_fn(x[..., :half]) * x[..., half:]
         return x