SwiGLU.Config(
                channels_in=4,
                channels_out=4,
                channels_hidden=4,
                expansion=2.6666666666666665,
                round_to=256,
                gate=True,
                bias=False,
                split_gate_projection=False,
                norm=None,
                act=<function priml.model.swiglu.silu at 0xdefacedeface>,
                depth_index=(),
                init_weight=<function priml.model.init.unit_fan_in_uniform at 0xdefacedeface>,
                init_weight_out=<function torch.nn.init.zeros_ at 0xdefacedeface>,
                shard=None
        )
