GatedDeltaNet.Config(
                channels_in=16,
                channels_out=16,
                num_heads_k=2,
                num_heads_v=2,
                channels_k_head=8,
                channels_v_head=8,
                conv_kernel_size=4,
                norm=CenteredRMSNorm.Config(
                        channels_in=8,
                        channels_out=8,
                        eps=1e-06
                ),
                init_weight=<function priml.model.init.kaiming_uniform at 0xdefacedeface>,
                init_conv_weight=<function priml.model.init.kaiming_uniform at 0xdefacedeface>,
                init_decay=<function priml.model.attention.gated_delta_net._init_decay at 0xdefacedeface>,
                depth_index=()
        )
