parakeet-rs 0.4.3

// Generated by scripts/generate_burn.py from encoder.onnx with burn-onnx 0.22.0. Do not edit.
extern crate alloc;
use burn::prelude::*;
use burn::nn::LayerNorm;
use burn::nn::LayerNormConfig;
use burn::nn::Linear;
use burn::nn::LinearConfig;
use burn::nn::PaddingConfig1d;
use burn::nn::PaddingConfig2d;
use burn::nn::conv::Conv1d;
use burn::nn::conv::Conv1dConfig;
use burn::nn::conv::Conv2d;
use burn::nn::conv::Conv2dConfig;


#[derive(Module, Debug)]
pub struct Submodule1 {
    conv2d1: Conv2d,
    constant675: burn::module::Param<Tensor<1, Int>>,
    conv2d2: Conv2d,
    conv2d3: Conv2d,
    conv2d4: Conv2d,
    conv2d5: Conv2d,
    linear1: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule1 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let conv2d1 = Conv2dConfig::new([1, 256], [3, 3])
            .with_stride([2, 2])
            .with_padding(PaddingConfig2d::Explicit(2, 2, 1, 1))
            .with_dilation([1, 1])
            .with_groups(1)
            .with_bias(true)
            .init(device);
        let constant675: burn::module::Param<Tensor<1, Int>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                1,
                Int,
            >::from_data(
                burn::tensor::TensorData::from([3i64]),
                (device, burn::tensor::DType::I64),
            ),
            device.clone(),
            false,
            [1].into(),
        );
        let conv2d2 = Conv2dConfig::new([256, 256], [3, 3])
            .with_stride([2, 2])
            .with_padding(PaddingConfig2d::Explicit(2, 2, 1, 1))
            .with_dilation([1, 1])
            .with_groups(256)
            .with_bias(true)
            .init(device);
        let conv2d3 = Conv2dConfig::new([256, 256], [1, 1])
            .with_stride([1, 1])
            .with_padding(PaddingConfig2d::Valid)
            .with_dilation([1, 1])
            .with_groups(1)
            .with_bias(true)
            .init(device);
        let conv2d4 = Conv2dConfig::new([256, 256], [3, 3])
            .with_stride([2, 2])
            .with_padding(PaddingConfig2d::Explicit(2, 2, 1, 1))
            .with_dilation([1, 1])
            .with_groups(256)
            .with_bias(true)
            .init(device);
        let conv2d5 = Conv2dConfig::new([256, 256], [1, 1])
            .with_stride([1, 1])
            .with_padding(PaddingConfig2d::Valid)
            .with_dilation([1, 1])
            .with_groups(1)
            .with_bias(true)
            .init(device);
        let linear1 = LinearConfig::new(4352, 1024).with_bias(true).init(device);
        Self {
            conv2d1,
            constant675,
            conv2d2,
            conv2d3,
            conv2d4,
            conv2d5,
            linear1,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        processed_signal: Tensor<3>,
        cache_last_channel: Tensor<4>,
        cache_last_time: Tensor<4>,
        processed_signal_length: Tensor<1, Int>,
    ) -> (Tensor<1, Int>, i64, i64, [i64; 1], bool, Tensor<4>, Tensor<4>, Tensor<3>) {
        let shape1_out1: [i64; 1] = {
            let axes = &processed_signal.clone().dims()[2..3];
            let mut output = [0i64; 1];
            for i in 0..1 {
                output[i] = axes[i] as i64;
            }
            output
        };
        let squeeze1_out1 = shape1_out1[0] as i64;
        let transpose1_out1 = cache_last_channel.permute([1, 0, 2, 3]);
        let constant670_out1 = 2i64;
        let div1_out1 = squeeze1_out1 / constant670_out1;
        let constant426_out1 = 1i64;
        let add1_out1 = constant426_out1 + div1_out1;
        let div2_out1 = add1_out1 / constant670_out1;
        let add2_out1 = constant426_out1 + div2_out1;
        let div3_out1 = add2_out1 / constant670_out1;
        let add3_out1 = constant426_out1 + div3_out1;
        let constant707_out1 = -1i64;
        let mul1_out1 = constant670_out1 * constant707_out1;
        let mul2_out1 = add3_out1 * constant707_out1;
        let max1_out1 = mul1_out1.max(mul2_out1);
        let mul3_out1 = max1_out1 * constant707_out1;
        let constant671_out1 = -1i64;
        let mul4_out1 = constant671_out1 * mul3_out1;
        let add4_out1 = constant426_out1 + mul4_out1;
        let add5_out1 = add4_out1 + div3_out1;
        let constant425_out1 = 0i64;
        let max2_out1 = constant425_out1.max(add5_out1);
        let mul5_out1 = constant670_out1 * max2_out1;
        let constant673_out1 = 139i64;
        let add6_out1 = constant673_out1 + mul5_out1;
        let transpose2_out1 = cache_last_time.permute([1, 0, 2, 3]);
        let transpose3_out1 = processed_signal.permute([0, 2, 1]);
        let unsqueeze1_out1: Tensor<4> = transpose3_out1.unsqueeze_dims::<4>(&[1]);
        let range1_out1 = {
            let start = 0i64;
            let delta = 1i64;
            assert!(delta != 0);
            let n = (((squeeze1_out1 as i128 - start as i128) as f64) / delta as f64)
                .ceil()
                .max(0.0) as i64;
            Tensor::arange(0..n, &self.device)
                .cast(burn::tensor::DType::I64)
                .mul_scalar(delta)
                .add_scalar(start)
        };
        let constant431_out1: [i64; 1] = [1i64];
        let concat1_out1: [i64; 2usize] = [&constant431_out1[..], &shape1_out1[..]]
            .concat()
            .try_into()
            .unwrap();
        let expand1_out1 = {
            let onnx_shape: [i64; 2usize] = concat1_out1;
            let input_dims = range1_out1.dims();
            let mut shape = onnx_shape;
            #[allow(clippy::needless_range_loop)]
            for i in 0..1usize {
                let dim_offset = 1usize + i;
                if shape[dim_offset] == 1 && input_dims[i] > 1 {
                    shape[dim_offset] = input_dims[i] as i64;
                }
            }
            range1_out1.expand(shape)
        };
        let unsqueeze2_out1: Tensor<2, Int> = processed_signal_length
            .clone()
            .unsqueeze_dims::<2>(&[1]);
        let less1_out1 = expand1_out1.lower(unsqueeze2_out1);
        let unsqueeze3_out1: Tensor<3, Bool> = less1_out1.unsqueeze_dims::<3>(&[-1]);
        let constant674_out1: [i64; 1] = [128i64];
        let concat2_out1: [i64; 3usize] = [
            &constant431_out1[..],
            &shape1_out1[..],
            &constant674_out1[..],
        ]
            .concat()
            .try_into()
            .unwrap();
        let expand2_out1 = {
            let onnx_shape: [i64; 3usize] = concat2_out1;
            let input_dims = unsqueeze3_out1.dims();
            let mut shape = onnx_shape;
            #[allow(clippy::needless_range_loop)]
            for i in 0..3usize {
                let dim_offset = i;
                if shape[dim_offset] == 1 && input_dims[i] > 1 {
                    shape[dim_offset] = input_dims[i] as i64;
                }
            }
            unsqueeze3_out1.expand(shape)
        };
        let cast1_out1 = expand2_out1.float().cast(burn::tensor::DType::F32);
        let unsqueeze4_out1: Tensor<4> = cast1_out1.unsqueeze_dims::<4>(&[1]);
        let mul6_out1 = unsqueeze1_out1.mul(unsqueeze4_out1);
        let conv2d1_out1 = self.conv2d1.forward(mul6_out1);
        let add7_out1 = processed_signal_length.add_scalar(constant670_out1);
        let add8_out1 = add7_out1.add_scalar(constant426_out1);
        let constant675_out1 = self.constant675.val();
        let sub1_out1 = add8_out1.sub(constant675_out1.clone());
        let less2_out1 = sub1_out1.clone().lower_elem(constant425_out1);
        let constant427_out1 = true;
        let equal1_out1 = less2_out1.equal_elem(constant427_out1);
        let mod1_out1 = sub1_out1.clone().remainder_scalar(constant670_out1);
        let cast2_out1 = mod1_out1.bool();
        let and1_out1 = equal1_out1.bool_and(cast2_out1);
        let cast3_out1 = and1_out1.int().cast(burn::tensor::DType::I64);
        let div4_out1 = sub1_out1.div_scalar(constant670_out1);
        let sub2_out1 = div4_out1.sub(cast3_out1);
        let add9_out1 = sub2_out1.add_scalar(constant426_out1);
        let range2_out1 = {
            let start = 0i64;
            let delta = 1i64;
            assert!(delta != 0);
            let n = (((add1_out1 as i128 - start as i128) as f64) / delta as f64)
                .ceil()
                .max(0.0) as i64;
            Tensor::arange(0..n, &self.device)
                .cast(burn::tensor::DType::I64)
                .mul_scalar(delta)
                .add_scalar(start)
        };
        let unsqueeze5_out1 = [add1_out1 as i64];
        let concat3_out1: [i64; 2usize] = [&constant431_out1[..], &unsqueeze5_out1[..]]
            .concat()
            .try_into()
            .unwrap();
        let expand3_out1 = {
            let onnx_shape: [i64; 2usize] = concat3_out1;
            let input_dims = range2_out1.dims();
            let mut shape = onnx_shape;
            #[allow(clippy::needless_range_loop)]
            for i in 0..1usize {
                let dim_offset = 1usize + i;
                if shape[dim_offset] == 1 && input_dims[i] > 1 {
                    shape[dim_offset] = input_dims[i] as i64;
                }
            }
            range2_out1.expand(shape)
        };
        let unsqueeze6_out1: Tensor<2, Int> = add9_out1
            .clone()
            .unsqueeze_dims::<2>(&[1]);
        let less3_out1 = expand3_out1.lower(unsqueeze6_out1);
        let unsqueeze7_out1: Tensor<3, Bool> = less3_out1.unsqueeze_dims::<3>(&[-1]);
        let constant676_out1: [i64; 1] = [65i64];
        let concat4_out1: [i64; 3usize] = [
            &constant431_out1[..],
            &unsqueeze5_out1[..],
            &constant676_out1[..],
        ]
            .concat()
            .try_into()
            .unwrap();
        let expand4_out1 = {
            let onnx_shape: [i64; 3usize] = concat4_out1;
            let input_dims = unsqueeze7_out1.dims();
            let mut shape = onnx_shape;
            #[allow(clippy::needless_range_loop)]
            for i in 0..3usize {
                let dim_offset = i;
                if shape[dim_offset] == 1 && input_dims[i] > 1 {
                    shape[dim_offset] = input_dims[i] as i64;
                }
            }
            unsqueeze7_out1.expand(shape)
        };
        let cast4_out1 = expand4_out1.float().cast(burn::tensor::DType::F32);
        let unsqueeze8_out1: Tensor<4> = cast4_out1.unsqueeze_dims::<4>(&[1]);
        let constant677_out1: [i64; 1] = [256i64];
        let concat5_out1: [i64; 4usize] = [
            &constant431_out1[..],
            &constant677_out1[..],
            &unsqueeze5_out1[..],
            &constant676_out1[..],
        ]
            .concat()
            .try_into()
            .unwrap();
        let expand5_out1 = {
            let onnx_shape: [i64; 4usize] = concat5_out1;
            let input_dims = unsqueeze8_out1.dims();
            let mut shape = onnx_shape;
            #[allow(clippy::needless_range_loop)]
            for i in 0..4usize {
                let dim_offset = i;
                if shape[dim_offset] == 1 && input_dims[i] > 1 {
                    shape[dim_offset] = input_dims[i] as i64;
                }
            }
            unsqueeze8_out1.expand(shape)
        };
        let mul7_out1 = conv2d1_out1.mul(expand5_out1.clone());
        let relu1_out1 = burn::tensor::activation::relu(mul7_out1);
        let mul8_out1 = relu1_out1.mul(expand5_out1);
        let conv2d2_out1 = self.conv2d2.forward(mul8_out1);
        let add10_out1 = add9_out1.add_scalar(constant670_out1);
        let add11_out1 = add10_out1.add_scalar(constant426_out1);
        let sub3_out1 = add11_out1.sub(constant675_out1.clone());
        let less4_out1 = sub3_out1.clone().lower_elem(constant425_out1);
        let equal2_out1 = less4_out1.equal_elem(constant427_out1);
        let mod2_out1 = sub3_out1.clone().remainder_scalar(constant670_out1);
        let cast5_out1 = mod2_out1.bool();
        let and2_out1 = equal2_out1.bool_and(cast5_out1);
        let cast6_out1 = and2_out1.int().cast(burn::tensor::DType::I64);
        let div5_out1 = sub3_out1.div_scalar(constant670_out1);
        let sub4_out1 = div5_out1.sub(cast6_out1);
        let add12_out1 = sub4_out1.add_scalar(constant426_out1);
        let range3_out1 = {
            let start = 0i64;
            let delta = 1i64;
            assert!(delta != 0);
            let n = (((add2_out1 as i128 - start as i128) as f64) / delta as f64)
                .ceil()
                .max(0.0) as i64;
            Tensor::arange(0..n, &self.device)
                .cast(burn::tensor::DType::I64)
                .mul_scalar(delta)
                .add_scalar(start)
        };
        let unsqueeze9_out1 = [add2_out1 as i64];
        let concat6_out1: [i64; 2usize] = [&constant431_out1[..], &unsqueeze9_out1[..]]
            .concat()
            .try_into()
            .unwrap();
        let expand6_out1 = {
            let onnx_shape: [i64; 2usize] = concat6_out1;
            let input_dims = range3_out1.dims();
            let mut shape = onnx_shape;
            #[allow(clippy::needless_range_loop)]
            for i in 0..1usize {
                let dim_offset = 1usize + i;
                if shape[dim_offset] == 1 && input_dims[i] > 1 {
                    shape[dim_offset] = input_dims[i] as i64;
                }
            }
            range3_out1.expand(shape)
        };
        let unsqueeze10_out1: Tensor<2, Int> = add12_out1
            .clone()
            .unsqueeze_dims::<2>(&[1]);
        let less5_out1 = expand6_out1.lower(unsqueeze10_out1);
        let unsqueeze11_out1: Tensor<3, Bool> = less5_out1.unsqueeze_dims::<3>(&[-1]);
        let constant678_out1: [i64; 1] = [33i64];
        let concat7_out1: [i64; 3usize] = [
            &constant431_out1[..],
            &unsqueeze9_out1[..],
            &constant678_out1[..],
        ]
            .concat()
            .try_into()
            .unwrap();
        let expand7_out1 = {
            let onnx_shape: [i64; 3usize] = concat7_out1;
            let input_dims = unsqueeze11_out1.dims();
            let mut shape = onnx_shape;
            #[allow(clippy::needless_range_loop)]
            for i in 0..3usize {
                let dim_offset = i;
                if shape[dim_offset] == 1 && input_dims[i] > 1 {
                    shape[dim_offset] = input_dims[i] as i64;
                }
            }
            unsqueeze11_out1.expand(shape)
        };
        let cast7_out1 = expand7_out1.float().cast(burn::tensor::DType::F32);
        let unsqueeze12_out1: Tensor<4> = cast7_out1.unsqueeze_dims::<4>(&[1]);
        let concat8_out1: [i64; 4usize] = [
            &constant431_out1[..],
            &constant677_out1[..],
            &unsqueeze9_out1[..],
            &constant678_out1[..],
        ]
            .concat()
            .try_into()
            .unwrap();
        let expand8_out1 = {
            let onnx_shape: [i64; 4usize] = concat8_out1;
            let input_dims = unsqueeze12_out1.dims();
            let mut shape = onnx_shape;
            #[allow(clippy::needless_range_loop)]
            for i in 0..4usize {
                let dim_offset = i;
                if shape[dim_offset] == 1 && input_dims[i] > 1 {
                    shape[dim_offset] = input_dims[i] as i64;
                }
            }
            unsqueeze12_out1.expand(shape)
        };
        let mul9_out1 = conv2d2_out1.mul(expand8_out1.clone());
        let conv2d3_out1 = self.conv2d3.forward(mul9_out1);
        let mul10_out1 = conv2d3_out1.mul(expand8_out1.clone());
        let relu2_out1 = burn::tensor::activation::relu(mul10_out1);
        let mul11_out1 = relu2_out1.mul(expand8_out1);
        let conv2d4_out1 = self.conv2d4.forward(mul11_out1);
        let add13_out1 = add12_out1.add_scalar(constant670_out1);
        let add14_out1 = add13_out1.add_scalar(constant426_out1);
        let sub5_out1 = add14_out1.sub(constant675_out1);
        let less6_out1 = sub5_out1.clone().lower_elem(constant425_out1);
        let equal3_out1 = less6_out1.equal_elem(constant427_out1);
        let mod3_out1 = sub5_out1.clone().remainder_scalar(constant670_out1);
        let cast8_out1 = mod3_out1.bool();
        let and3_out1 = equal3_out1.bool_and(cast8_out1);
        let cast9_out1 = and3_out1.int().cast(burn::tensor::DType::I64);
        let div6_out1 = sub5_out1.div_scalar(constant670_out1);
        let sub6_out1 = div6_out1.sub(cast9_out1);
        let add15_out1 = sub6_out1.add_scalar(constant426_out1);
        let range4_out1 = {
            let start = 0i64;
            let delta = 1i64;
            assert!(delta != 0);
            let n = (((add3_out1 as i128 - start as i128) as f64) / delta as f64)
                .ceil()
                .max(0.0) as i64;
            Tensor::arange(0..n, &self.device)
                .cast(burn::tensor::DType::I64)
                .mul_scalar(delta)
                .add_scalar(start)
        };
        let unsqueeze13_out1 = [add3_out1 as i64];
        let concat9_out1: [i64; 2usize] = [&constant431_out1[..], &unsqueeze13_out1[..]]
            .concat()
            .try_into()
            .unwrap();
        let expand9_out1 = {
            let onnx_shape: [i64; 2usize] = concat9_out1;
            let input_dims = range4_out1.dims();
            let mut shape = onnx_shape;
            #[allow(clippy::needless_range_loop)]
            for i in 0..1usize {
                let dim_offset = 1usize + i;
                if shape[dim_offset] == 1 && input_dims[i] > 1 {
                    shape[dim_offset] = input_dims[i] as i64;
                }
            }
            range4_out1.expand(shape)
        };
        let unsqueeze14_out1: Tensor<2, Int> = add15_out1
            .clone()
            .unsqueeze_dims::<2>(&[1]);
        let less7_out1 = expand9_out1.lower(unsqueeze14_out1);
        let unsqueeze15_out1: Tensor<3, Bool> = less7_out1.unsqueeze_dims::<3>(&[-1]);
        let constant679_out1: [i64; 1] = [17i64];
        let concat10_out1: [i64; 3usize] = [
            &constant431_out1[..],
            &unsqueeze13_out1[..],
            &constant679_out1[..],
        ]
            .concat()
            .try_into()
            .unwrap();
        let expand10_out1 = {
            let onnx_shape: [i64; 3usize] = concat10_out1;
            let input_dims = unsqueeze15_out1.dims();
            let mut shape = onnx_shape;
            #[allow(clippy::needless_range_loop)]
            for i in 0..3usize {
                let dim_offset = i;
                if shape[dim_offset] == 1 && input_dims[i] > 1 {
                    shape[dim_offset] = input_dims[i] as i64;
                }
            }
            unsqueeze15_out1.expand(shape)
        };
        let cast10_out1 = expand10_out1.float().cast(burn::tensor::DType::F32);
        let unsqueeze16_out1: Tensor<4> = cast10_out1.unsqueeze_dims::<4>(&[1]);
        let concat11_out1: [i64; 4usize] = [
            &constant431_out1[..],
            &constant677_out1[..],
            &unsqueeze13_out1[..],
            &constant679_out1[..],
        ]
            .concat()
            .try_into()
            .unwrap();
        let expand11_out1 = {
            let onnx_shape: [i64; 4usize] = concat11_out1;
            let input_dims = unsqueeze16_out1.dims();
            let mut shape = onnx_shape;
            #[allow(clippy::needless_range_loop)]
            for i in 0..4usize {
                let dim_offset = i;
                if shape[dim_offset] == 1 && input_dims[i] > 1 {
                    shape[dim_offset] = input_dims[i] as i64;
                }
            }
            unsqueeze16_out1.expand(shape)
        };
        let mul12_out1 = conv2d4_out1.mul(expand11_out1.clone());
        let conv2d5_out1 = self.conv2d5.forward(mul12_out1);
        let mul13_out1 = conv2d5_out1.mul(expand11_out1.clone());
        let relu3_out1 = burn::tensor::activation::relu(mul13_out1);
        let mul14_out1 = relu3_out1.mul(expand11_out1);
        let transpose4_out1 = mul14_out1.permute([0, 2, 1, 3]);
        let reshape1_out1 = transpose4_out1.reshape([1, -1, 4352]);
        let linear1_out1 = self.linear1.forward(reshape1_out1);
        let slice1_out1 = linear1_out1.slice(s![.., 2.., ..]);
        let sub7_out1 = add15_out1.sub_scalar(constant670_out1);
        (
            sub7_out1,
            max2_out1,
            add6_out1,
            constant431_out1,
            constant427_out1,
            transpose1_out1,
            transpose2_out1,
            slice1_out1,
        )
    }
}
#[derive(Module, Debug)]
pub struct Submodule2 {
    constant711: burn::module::Param<Tensor<1, Int>>,
    constant424: burn::module::Param<Tensor<3>>,
    constant434: burn::module::Param<Tensor<1>>,
    linear2: Linear,
    linear3: Linear,
    linear4: Linear,
    linear5: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule2 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let constant711: burn::module::Param<Tensor<1, Int>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                1,
                Int,
            >::from_data(
                burn::tensor::TensorData::from([-1i64]),
                (device, burn::tensor::DType::I64),
            ),
            device.clone(),
            false,
            [1].into(),
        );
        let constant424: burn::module::Param<Tensor<3>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                3,
            >::zeros([1, 9999, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1, 9999, 1024].into(),
        );
        let constant434: burn::module::Param<Tensor<1>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                1,
            >::from_data(
                burn::tensor::TensorData::from([14f64]),
                (device, burn::tensor::DType::F32),
            ),
            device.clone(),
            false,
            [1].into(),
        );
        let linear2 = LinearConfig::new(1024, 1024).with_bias(true).init(device);
        let linear3 = LinearConfig::new(1024, 1024).with_bias(true).init(device);
        let linear4 = LinearConfig::new(1024, 1024).with_bias(true).init(device);
        let linear5 = LinearConfig::new(1024, 1024).with_bias(true).init(device);
        Self {
            constant711,
            constant424,
            constant434,
            linear2,
            linear3,
            linear4,
            linear5,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        sub7_out1: Tensor<1, Int>,
        max2_out1: i64,
        cache_last_channel_len: Tensor<1, Int>,
        add6_out1: i64,
        constant431_out1: [i64; 1],
        constant427_out1: bool,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        spk_targets: Tensor<2>,
        slice1_out1: Tensor<3>,
        bg_spk_targets: Tensor<2>,
    ) -> (
        Tensor<3>,
        Tensor<3>,
        Tensor<3>,
        [i64; 1],
        Tensor<3, Bool>,
        Tensor<2, Bool>,
        Tensor<3>,
        Tensor<1, Int>,
    ) {
        let clip1_out1 = sub7_out1.clamp_min(0f64);
        let constant665_out1 = 70i64;
        let add16_out1 = max2_out1 + constant665_out1;
        let add17_out1 = clip1_out1.clone().add_scalar(constant665_out1);
        let constant711_out1 = self.constant711.val();
        let mul15_out1 = cache_last_channel_len.mul(constant711_out1);
        let add18_out1 = mul15_out1.add_scalar(constant665_out1);
        let constant681_out1 = 5000i64;
        let sub8_out1 = constant681_out1 - add16_out1;
        let add19_out1 = sub8_out1 + add6_out1;
        let unsqueeze17_out1 = [sub8_out1 as i64];
        let unsqueeze18_out1 = [add19_out1 as i64];
        let constant424_out1 = self.constant424.val();
        let transpose5_out1 = constant424_out1.permute([1, 0, 2]);
        let slice2_out1 = transpose5_out1
            .slice(s![unsqueeze17_out1[0]..unsqueeze18_out1[0], .., ..]);
        let transpose6_out1 = slice2_out1.permute([1, 0, 2]);
        let unsqueeze19_out1 = [add16_out1 as i64];
        let concat12_out1: [i64; 3usize] = [
            &constant431_out1[..],
            &unsqueeze19_out1[..],
            &unsqueeze19_out1[..],
        ]
            .concat()
            .try_into()
            .unwrap();
        let expand12_out1 = {
            let input = Tensor::<
                1,
                Bool,
            >::from_data(
                burn::tensor::TensorData::from([constant427_out1]),
                (
                    &self.device,
                    burn::tensor::DType::Bool(burn::tensor::BoolStore::Native),
                ),
            );
            input.expand(concat12_out1)
        };
        let cast11_out1 = add16_out1;
        let range5_out1 = {
            let start = 0i32;
            let delta = 1i32;
            assert!(delta != 0);
            let n = (((cast11_out1 as i128 - start as i128) as f64) / delta as f64)
                .ceil()
                .max(0.0) as i64;
            Tensor::arange(0..n, &self.device)
                .cast(burn::tensor::DType::I32)
                .mul_scalar(delta)
                .add_scalar(start)
        };
        let cast12_out1 = range5_out1.float().cast(burn::tensor::DType::F32);
        let constant434_out1 = self.constant434.val();
        let div7_out1 = cast12_out1.div(constant434_out1);
        let abs1_out1 = div7_out1.clone().abs();
        let floor1_out1 = abs1_out1.floor();
        let sign1_out1 = div7_out1.sign();
        let mul16_out1 = floor1_out1.mul(sign1_out1);
        let cast13_out1 = mul16_out1.int().cast(burn::tensor::DType::I32);
        let unsqueeze20_out1: Tensor<2, Int> = cast13_out1
            .clone()
            .unsqueeze_dims::<2>(&[1]);
        let unsqueeze21_out1: Tensor<2, Int> = cast13_out1.unsqueeze_dims::<2>(&[0]);
        let sub9_out1 = unsqueeze20_out1.sub(unsqueeze21_out1);
        let constant435_out1 = 5i32;
        let lessorequal1_out1 = sub9_out1.clone().lower_equal_elem(constant435_out1);
        let constant432_out1 = 0i32;
        let greaterorequal1_out1 = sub9_out1.greater_equal_elem(constant432_out1);
        let and4_out1 = lessorequal1_out1.bool_and(greaterorequal1_out1);
        let unsqueeze22_out1: Tensor<3, Bool> = and4_out1.unsqueeze_dims::<3>(&[0]);
        let and5_out1 = expand12_out1.bool_and(unsqueeze22_out1);
        let range6_out1 = {
            let start = 0i64;
            let delta = 1i64;
            assert!(delta != 0);
            let n = (((add16_out1 as i128 - start as i128) as f64) / delta as f64)
                .ceil()
                .max(0.0) as i64;
            Tensor::arange(0..n, &self.device)
                .cast(burn::tensor::DType::I64)
                .mul_scalar(delta)
                .add_scalar(start)
        };
        let expand13_out1 = {
            let onnx_shape: [i64; 2usize] = [1, 1];
            let input_dims = range6_out1.dims();
            let mut shape = onnx_shape;
            #[allow(clippy::needless_range_loop)]
            for i in 0..1usize {
                let dim_offset = 1usize + i;
                if shape[dim_offset] == 1 && input_dims[i] > 1 {
                    shape[dim_offset] = input_dims[i] as i64;
                }
            }
            range6_out1.expand(shape)
        };
        let unsqueeze23_out1: Tensor<2, Int> = add17_out1.unsqueeze_dims::<2>(&[-1]);
        let less8_out1 = expand13_out1.clone().lower(unsqueeze23_out1);
        let unsqueeze24_out1: Tensor<2, Int> = add18_out1.unsqueeze_dims::<2>(&[-1]);
        let greaterorequal2_out1 = expand13_out1.greater_equal(unsqueeze24_out1);
        let and6_out1 = greaterorequal2_out1.bool_and(less8_out1);
        let unsqueeze25_out1: Tensor<3, Bool> = and6_out1
            .clone()
            .unsqueeze_dims::<3>(&[1]);
        let expand14_out1 = {
            let onnx_shape: [i64; 3usize] = [1, 1, 1];
            let input_dims = unsqueeze25_out1.dims();
            let mut shape = onnx_shape;
            #[allow(clippy::needless_range_loop)]
            for i in 0..3usize {
                let dim_offset = i;
                if shape[dim_offset] == 1 && input_dims[i] > 1 {
                    shape[dim_offset] = input_dims[i] as i64;
                }
            }
            unsqueeze25_out1.expand(shape)
        };
        let concat13_out1: [i64; 3usize] = [
            &constant431_out1[..],
            &unsqueeze19_out1[..],
            &constant431_out1[..],
        ]
            .concat()
            .try_into()
            .unwrap();
        let tile1_out1 = {
            let repeats: alloc::vec::Vec<usize> = concat13_out1
                .iter()
                .map(|&v| v as usize)
                .collect();
            expand14_out1.repeat(&repeats)
        };
        let transpose7_out1 = tile1_out1.clone().permute([0, 2, 1]);
        let and7_out1 = tile1_out1.bool_and(transpose7_out1);
        let and8_out1 = and7_out1.bool_and(and5_out1);
        let not1_out1 = and8_out1.bool_not();
        let not2_out1 = and6_out1.bool_not();
        let slice3_out1 = not2_out1.slice(s![.., 70..]);
        let slice4_out1 = not1_out1.slice(s![.., 70.., ..]);
        let gather1_out1 = {
            let sliced = transpose1_out1.slice(s![0, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather2_out1 = {
            let sliced = transpose2_out1.slice(s![0, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let constant716_out1 = -1i64;
        let mul17_out1 = max2_out1 * constant716_out1;
        let unsqueeze26_out1 = [mul17_out1 as i64];
        let transpose8_out1 = spk_targets.permute([1, 0]);
        let slice5_out1 = transpose8_out1
            .slice(s![unsqueeze26_out1[0]..9223372036854775807, ..]);
        let transpose9_out1 = slice5_out1.permute([1, 0]);
        let unsqueeze27_out1: Tensor<3> = transpose9_out1.unsqueeze_dims::<3>(&[2]);
        let mul18_out1 = slice1_out1.clone().mul(unsqueeze27_out1);
        let linear2_out1 = self.linear2.forward(mul18_out1);
        let relu4_out1 = burn::tensor::activation::relu(linear2_out1);
        let linear3_out1 = self.linear3.forward(relu4_out1);
        let add20_out1 = slice1_out1.add(linear3_out1);
        let transpose10_out1 = bg_spk_targets.permute([1, 0]);
        let slice6_out1 = transpose10_out1
            .slice(s![unsqueeze26_out1[0]..9223372036854775807, ..]);
        let transpose11_out1 = slice6_out1.permute([1, 0]);
        let unsqueeze28_out1: Tensor<3> = transpose11_out1.unsqueeze_dims::<3>(&[2]);
        let mul19_out1 = add20_out1.clone().mul(unsqueeze28_out1);
        let linear4_out1 = self.linear4.forward(mul19_out1);
        let relu5_out1 = burn::tensor::activation::relu(linear4_out1);
        let linear5_out1 = self.linear5.forward(relu5_out1);
        let add21_out1 = add20_out1.add(linear5_out1);
        (
            add21_out1,
            gather1_out1,
            transpose6_out1,
            unsqueeze19_out1,
            slice4_out1,
            slice3_out1,
            gather2_out1,
            clip1_out1,
        )
    }
}
#[derive(Module, Debug)]
pub struct Submodule3 {
    layernormalization1: LayerNorm,
    linear6: Linear,
    linear7: Linear,
    constant684: burn::module::Param<Tensor<1>>,
    layernormalization2: LayerNorm,
    linear8: Linear,
    linear9: Linear,
    linear10: Linear,
    linear11: Linear,
    constant23: burn::module::Param<Tensor<2>>,
    constant24: burn::module::Param<Tensor<2>>,
    constant687: burn::module::Param<Tensor<1>>,
    linear12: Linear,
    layernormalization3: LayerNorm,
    squeeze2: burn::module::Param<Tensor<2>>,
    #[module(skip)]
    device: Device,
}
impl Submodule3 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization1 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear6 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear7 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let constant684: burn::module::Param<Tensor<1>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                1,
            >::from_data(
                burn::tensor::TensorData::from([0.5f64]),
                (device, burn::tensor::DType::F32),
            ),
            device.clone(),
            false,
            [1].into(),
        );
        let layernormalization2 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear8 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear9 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear10 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear11 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant23: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant24: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant687: burn::module::Param<Tensor<1>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                1,
            >::from_data(
                burn::tensor::TensorData::from([0.0883883461356163f64]),
                (device, burn::tensor::DType::F32),
            ),
            device.clone(),
            false,
            [1].into(),
        );
        let linear12 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let layernormalization3 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze2: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        Self {
            layernormalization1,
            linear6,
            linear7,
            constant684,
            layernormalization2,
            linear8,
            linear9,
            linear10,
            linear11,
            constant23,
            constant24,
            constant687,
            linear12,
            layernormalization3,
            squeeze2,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add21_out1: Tensor<3>,
        gather1_out1: Tensor<3>,
        max2_out1: i64,
        transpose6_out1: Tensor<3>,
        constant431_out1: [i64; 1],
        add6_out1: i64,
        unsqueeze19_out1: [i64; 1],
        slice4_out1: Tensor<3, Bool>,
        slice3_out1: Tensor<2, Bool>,
        gather2_out1: Tensor<3>,
    ) -> (
        Tensor<3>,
        Tensor<3>,
        Tensor<1>,
        [i64; 1],
        [i64; 4],
        [i64; 4],
        Tensor<1>,
        Tensor<4, Bool>,
        f32,
        Tensor<4, Bool>,
        f32,
        Tensor<3, Bool>,
        Tensor<3>,
    ) {
        let layernormalization1_out1 = {
            let dtype = add21_out1.clone().dtype();
            self.layernormalization1
                .forward(add21_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear6_out1 = self.linear6.forward(layernormalization1_out1);
        let sigmoid1_out1 = burn::tensor::activation::sigmoid(linear6_out1.clone());
        let mul20_out1 = linear6_out1.mul(sigmoid1_out1);
        let linear7_out1 = self.linear7.forward(mul20_out1);
        let constant684_out1 = self.constant684.val();
        let mul21_out1 = linear7_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add22_out1 = add21_out1.add(mul21_out1);
        let layernormalization2_out1 = {
            let dtype = add22_out1.clone().dtype();
            self.layernormalization2
                .forward(add22_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat14_out1 = burn::tensor::Tensor::cat(
            [gather1_out1.clone(), layernormalization2_out1.clone()].into(),
            1,
        );
        let unsqueeze29_out1 = [max2_out1 as i64];
        let transpose12_out1 = gather1_out1.permute([1, 0, 2]);
        let slice7_out1 = transpose12_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose13_out1 = slice7_out1.permute([1, 0, 2]);
        let concat15_out1 = burn::tensor::Tensor::cat(
            [transpose13_out1, layernormalization2_out1.clone()].into(),
            1,
        );
        let linear8_out1 = self.linear8.forward(layernormalization2_out1);
        let reshape2_out1 = linear8_out1.reshape([1, -1, 8, 128]);
        let linear9_out1 = self.linear9.forward(concat14_out1.clone());
        let reshape3_out1 = linear9_out1.reshape([1, -1, 8, 128]);
        let linear10_out1 = self.linear10.forward(concat14_out1);
        let reshape4_out1 = linear10_out1.reshape([1, -1, 8, 128]);
        let transpose14_out1 = reshape3_out1.permute([0, 2, 1, 3]);
        let transpose15_out1 = reshape4_out1.permute([0, 2, 1, 3]);
        let linear11_out1 = self.linear11.forward(transpose6_out1);
        let reshape5_out1 = linear11_out1.reshape([1, -1, 8, 128]);
        let constant23_out1 = self.constant23.val();
        let add23_out1 = reshape2_out1
            .clone()
            .add((constant23_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose16_out1 = add23_out1.permute([0, 2, 1, 3]);
        let constant24_out1 = self.constant24.val();
        let add24_out1 = reshape2_out1
            .add((constant24_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose17_out1 = add24_out1.permute([0, 2, 1, 3]);
        let transpose18_out1 = reshape5_out1.permute([0, 2, 3, 1]);
        let matmul12_out1 = transpose17_out1.matmul(transpose18_out1);
        let pad1_out1 = matmul12_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let constant451_out1: [i64; 1] = [-1i64];
        let constant686_out1: [i64; 1] = [8i64];
        let concat16_out1: [i64; 4usize] = [
            &constant431_out1[..],
            &constant686_out1[..],
            &constant451_out1[..],
            &unsqueeze29_out1[..],
        ]
            .concat()
            .try_into()
            .unwrap();
        let reshape6_out1 = pad1_out1.reshape(concat16_out1);
        let slice8_out1 = reshape6_out1.slice(s![.., .., 1.., ..]);
        let unsqueeze30_out1 = [add6_out1 as i64];
        let concat17_out1: [i64; 4usize] = [
            &constant431_out1[..],
            &constant686_out1[..],
            &unsqueeze29_out1[..],
            &unsqueeze30_out1[..],
        ]
            .concat()
            .try_into()
            .unwrap();
        let reshape7_out1 = slice8_out1.reshape(concat17_out1);
        let transpose19_out1 = reshape7_out1.permute([3, 0, 1, 2]);
        let slice9_out1 = transpose19_out1.slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose20_out1 = slice9_out1.permute([1, 2, 3, 0]);
        let constant687_out1 = self.constant687.val();
        let mul22_out1 = transpose20_out1
            .mul((constant687_out1.clone()).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let unsqueeze31_out1: Tensor<4, Bool> = slice4_out1.unsqueeze_dims::<4>(&[1]);
        let constant450_out1 = -10000f32;
        let where1_out1 = mul22_out1
            .mask_fill(unsqueeze31_out1.clone(), constant450_out1);
        let (attention1_out1,) = {
            let q = transpose16_out1;
            let k = transpose14_out1;
            let v = transpose15_out1;
            let score_bias = where1_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention1_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention1_out1 = {
                let dims = attention1_out1.dims();
                attention1_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention1_out1,)
        };
        let cast14_out1 = unsqueeze31_out1.clone().int().cast(burn::tensor::DType::I64);
        let reducemin1_out1 = {
            cast14_out1.min_dim(3usize).squeeze_dims::<3usize>(&[3])
        };
        let cast15_out1 = reducemin1_out1.bool();
        let unsqueeze32_out1: Tensor<4, Bool> = cast15_out1.unsqueeze_dims::<4>(&[-1]);
        let expand15_out1 = {
            let onnx_shape: [i64; 4usize] = [1, 8, 1, 128];
            let input_dims = unsqueeze32_out1.dims();
            let mut shape = onnx_shape;
            #[allow(clippy::needless_range_loop)]
            for i in 0..4usize {
                let dim_offset = i;
                if shape[dim_offset] == 1 && input_dims[i] > 1 {
                    shape[dim_offset] = input_dims[i] as i64;
                }
            }
            unsqueeze32_out1.expand(shape)
        };
        let constant453_out1 = 0f32;
        let where2_out1 = attention1_out1
            .mask_fill(expand15_out1.clone(), constant453_out1);
        let transpose21_out1 = where2_out1.permute([0, 2, 1, 3]);
        let reshape8_out1 = transpose21_out1.reshape([1, -1, 1024]);
        let linear12_out1 = self.linear12.forward(reshape8_out1);
        let add25_out1 = add22_out1.add(linear12_out1);
        let layernormalization3_out1 = {
            let dtype = add25_out1.clone().dtype();
            self.layernormalization3
                .forward(add25_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze2_out1 = self.squeeze2.val();
        let transpose22_out1 = squeeze2_out1.permute([1, 0]);
        let matmul14_out1 = layernormalization3_out1
            .matmul(transpose22_out1.unsqueeze::<3usize>());
        let split_tensors = matmul14_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split1_out1, split1_out2] = split_tensors.try_into().unwrap();
        let sigmoid2_out1 = burn::tensor::activation::sigmoid(split1_out2);
        let mul23_out1 = split1_out1.mul(sigmoid2_out1);
        let transpose23_out1 = mul23_out1.permute([0, 2, 1]);
        let unsqueeze33_out1: Tensor<3, Bool> = slice3_out1.unsqueeze_dims::<3>(&[1]);
        let where3_out1 = transpose23_out1
            .mask_fill(unsqueeze33_out1.clone(), constant453_out1);
        let pad2_out1 = where3_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat18_out1 = burn::tensor::Tensor::cat(
            [gather2_out1, pad2_out1].into(),
            2,
        );
        (
            concat18_out1,
            add25_out1,
            constant684_out1,
            unsqueeze29_out1,
            concat16_out1,
            concat17_out1,
            constant687_out1,
            unsqueeze31_out1,
            constant450_out1,
            expand15_out1,
            constant453_out1,
            unsqueeze33_out1,
            concat15_out1,
        )
    }
}
#[derive(Module, Debug)]
pub struct Submodule4 {
    conv1d1: Conv1d,
    layernormalization4: LayerNorm,
    squeeze3: burn::module::Param<Tensor<2>>,
    layernormalization5: LayerNorm,
    linear13: Linear,
    linear14: Linear,
    layernormalization6: LayerNorm,
    layernormalization7: LayerNorm,
    linear15: Linear,
    linear16: Linear,
    layernormalization8: LayerNorm,
    linear17: Linear,
    linear18: Linear,
    linear19: Linear,
    linear20: Linear,
    constant40: burn::module::Param<Tensor<2>>,
    constant41: burn::module::Param<Tensor<2>>,
    linear21: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule4 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let conv1d1 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization4 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze3: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization5 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear13 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear14 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization6 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization7 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear15 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear16 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization8 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear17 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear18 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear19 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear20 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant40: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant41: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear21 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            conv1d1,
            layernormalization4,
            squeeze3,
            layernormalization5,
            linear13,
            linear14,
            layernormalization6,
            layernormalization7,
            linear15,
            linear16,
            layernormalization8,
            linear17,
            linear18,
            linear19,
            linear20,
            constant40,
            constant41,
            linear21,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        concat18_out1: Tensor<3>,
        add25_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
        constant453_out1: f32,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let slice10_out1 = concat18_out1.clone().slice(s![.., .., - 8..]);
        let conv1d1_out1 = self.conv1d1.forward(concat18_out1);
        let transpose24_out1 = conv1d1_out1.permute([0, 2, 1]);
        let layernormalization4_out1 = {
            let dtype = transpose24_out1.dtype();
            self.layernormalization4
                .forward(transpose24_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid3_out1 = burn::tensor::activation::sigmoid(
            layernormalization4_out1.clone(),
        );
        let mul24_out1 = layernormalization4_out1.mul(sigmoid3_out1);
        let squeeze3_out1 = self.squeeze3.val();
        let transpose25_out1 = squeeze3_out1.permute([1, 0]);
        let matmul15_out1 = mul24_out1.matmul(transpose25_out1.unsqueeze::<3usize>());
        let add26_out1 = add25_out1.add(matmul15_out1);
        let layernormalization5_out1 = {
            let dtype = add26_out1.clone().dtype();
            self.layernormalization5
                .forward(add26_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear13_out1 = self.linear13.forward(layernormalization5_out1);
        let sigmoid4_out1 = burn::tensor::activation::sigmoid(linear13_out1.clone());
        let mul25_out1 = linear13_out1.mul(sigmoid4_out1);
        let linear14_out1 = self.linear14.forward(mul25_out1);
        let mul26_out1 = linear14_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add27_out1 = add26_out1.add(mul26_out1);
        let layernormalization6_out1 = {
            let dtype = add27_out1.dtype();
            self.layernormalization6
                .forward(add27_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather3_out1 = {
            let sliced = transpose1_out1.slice(s![1, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather4_out1 = {
            let sliced = transpose2_out1.slice(s![1, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization7_out1 = {
            let dtype = layernormalization6_out1.clone().dtype();
            self.layernormalization7
                .forward(layernormalization6_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear15_out1 = self.linear15.forward(layernormalization7_out1);
        let sigmoid5_out1 = burn::tensor::activation::sigmoid(linear15_out1.clone());
        let mul27_out1 = linear15_out1.mul(sigmoid5_out1);
        let linear16_out1 = self.linear16.forward(mul27_out1);
        let mul28_out1 = linear16_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add28_out1 = layernormalization6_out1.add(mul28_out1);
        let layernormalization8_out1 = {
            let dtype = add28_out1.clone().dtype();
            self.layernormalization8
                .forward(add28_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat19_out1 = burn::tensor::Tensor::cat(
            [gather3_out1.clone(), layernormalization8_out1.clone()].into(),
            1,
        );
        let transpose26_out1 = gather3_out1.permute([1, 0, 2]);
        let slice11_out1 = transpose26_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose27_out1 = slice11_out1.permute([1, 0, 2]);
        let concat20_out1 = burn::tensor::Tensor::cat(
            [transpose27_out1, layernormalization8_out1.clone()].into(),
            1,
        );
        let linear17_out1 = self.linear17.forward(layernormalization8_out1);
        let reshape9_out1 = linear17_out1.reshape([1, -1, 8, 128]);
        let linear18_out1 = self.linear18.forward(concat19_out1.clone());
        let reshape10_out1 = linear18_out1.reshape([1, -1, 8, 128]);
        let linear19_out1 = self.linear19.forward(concat19_out1);
        let reshape11_out1 = linear19_out1.reshape([1, -1, 8, 128]);
        let transpose28_out1 = reshape10_out1.permute([0, 2, 1, 3]);
        let transpose29_out1 = reshape11_out1.permute([0, 2, 1, 3]);
        let linear20_out1 = self.linear20.forward(transpose6_out1);
        let reshape12_out1 = linear20_out1.reshape([1, -1, 8, 128]);
        let constant40_out1 = self.constant40.val();
        let add29_out1 = reshape9_out1
            .clone()
            .add((constant40_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose30_out1 = add29_out1.permute([0, 2, 1, 3]);
        let constant41_out1 = self.constant41.val();
        let add30_out1 = reshape9_out1
            .add((constant41_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose31_out1 = add30_out1.permute([0, 2, 1, 3]);
        let transpose32_out1 = reshape12_out1.permute([0, 2, 3, 1]);
        let matmul24_out1 = transpose31_out1.matmul(transpose32_out1);
        let pad3_out1 = matmul24_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape13_out1 = pad3_out1.reshape(concat16_out1);
        let slice12_out1 = reshape13_out1.slice(s![.., .., 1.., ..]);
        let reshape14_out1 = slice12_out1.reshape(concat17_out1);
        let transpose33_out1 = reshape14_out1.permute([3, 0, 1, 2]);
        let slice13_out1 = transpose33_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose34_out1 = slice13_out1.permute([1, 2, 3, 0]);
        let mul29_out1 = transpose34_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where4_out1 = mul29_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention2_out1,) = {
            let q = transpose30_out1;
            let k = transpose28_out1;
            let v = transpose29_out1;
            let score_bias = where4_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention2_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention2_out1 = {
                let dims = attention2_out1.dims();
                attention2_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention2_out1,)
        };
        let where5_out1 = attention2_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose35_out1 = where5_out1.permute([0, 2, 1, 3]);
        let reshape15_out1 = transpose35_out1.reshape([1, -1, 1024]);
        let linear21_out1 = self.linear21.forward(reshape15_out1);
        let add31_out1 = add28_out1.add(linear21_out1);
        (add31_out1, gather4_out1, concat20_out1, slice10_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule5 {
    layernormalization9: LayerNorm,
    squeeze4: burn::module::Param<Tensor<2>>,
    conv1d2: Conv1d,
    layernormalization10: LayerNorm,
    squeeze5: burn::module::Param<Tensor<2>>,
    layernormalization11: LayerNorm,
    linear22: Linear,
    linear23: Linear,
    layernormalization12: LayerNorm,
    layernormalization13: LayerNorm,
    linear24: Linear,
    linear25: Linear,
    layernormalization14: LayerNorm,
    linear26: Linear,
    linear27: Linear,
    linear28: Linear,
    linear29: Linear,
    constant57: burn::module::Param<Tensor<2>>,
    constant58: burn::module::Param<Tensor<2>>,
    linear30: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule5 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization9 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze4: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d2 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization10 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze5: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization11 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear22 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear23 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization12 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization13 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear24 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear25 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization14 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear26 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear27 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear28 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear29 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant57: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant58: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear30 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization9,
            squeeze4,
            conv1d2,
            layernormalization10,
            squeeze5,
            layernormalization11,
            linear22,
            linear23,
            layernormalization12,
            layernormalization13,
            linear24,
            linear25,
            layernormalization14,
            linear26,
            linear27,
            linear28,
            linear29,
            constant57,
            constant58,
            linear30,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add31_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather4_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization9_out1 = {
            let dtype = add31_out1.clone().dtype();
            self.layernormalization9
                .forward(add31_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze4_out1 = self.squeeze4.val();
        let transpose36_out1 = squeeze4_out1.permute([1, 0]);
        let matmul26_out1 = layernormalization9_out1
            .matmul(transpose36_out1.unsqueeze::<3usize>());
        let split_tensors = matmul26_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split2_out1, split2_out2] = split_tensors.try_into().unwrap();
        let sigmoid6_out1 = burn::tensor::activation::sigmoid(split2_out2);
        let mul30_out1 = split2_out1.mul(sigmoid6_out1);
        let transpose37_out1 = mul30_out1.permute([0, 2, 1]);
        let where6_out1 = transpose37_out1.mask_fill(unsqueeze33_out1, constant453_out1);
        let pad4_out1 = where6_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat21_out1 = burn::tensor::Tensor::cat(
            [gather4_out1, pad4_out1].into(),
            2,
        );
        let slice14_out1 = concat21_out1.clone().slice(s![.., .., - 8..]);
        let conv1d2_out1 = self.conv1d2.forward(concat21_out1);
        let transpose38_out1 = conv1d2_out1.permute([0, 2, 1]);
        let layernormalization10_out1 = {
            let dtype = transpose38_out1.dtype();
            self.layernormalization10
                .forward(transpose38_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid7_out1 = burn::tensor::activation::sigmoid(
            layernormalization10_out1.clone(),
        );
        let mul31_out1 = layernormalization10_out1.mul(sigmoid7_out1);
        let squeeze5_out1 = self.squeeze5.val();
        let transpose39_out1 = squeeze5_out1.permute([1, 0]);
        let matmul27_out1 = mul31_out1.matmul(transpose39_out1.unsqueeze::<3usize>());
        let add32_out1 = add31_out1.add(matmul27_out1);
        let layernormalization11_out1 = {
            let dtype = add32_out1.clone().dtype();
            self.layernormalization11
                .forward(add32_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear22_out1 = self.linear22.forward(layernormalization11_out1);
        let sigmoid8_out1 = burn::tensor::activation::sigmoid(linear22_out1.clone());
        let mul32_out1 = linear22_out1.mul(sigmoid8_out1);
        let linear23_out1 = self.linear23.forward(mul32_out1);
        let mul33_out1 = linear23_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add33_out1 = add32_out1.add(mul33_out1);
        let layernormalization12_out1 = {
            let dtype = add33_out1.dtype();
            self.layernormalization12
                .forward(add33_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather5_out1 = {
            let sliced = transpose1_out1.slice(s![2, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather6_out1 = {
            let sliced = transpose2_out1.slice(s![2, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization13_out1 = {
            let dtype = layernormalization12_out1.clone().dtype();
            self.layernormalization13
                .forward(
                    layernormalization12_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear24_out1 = self.linear24.forward(layernormalization13_out1);
        let sigmoid9_out1 = burn::tensor::activation::sigmoid(linear24_out1.clone());
        let mul34_out1 = linear24_out1.mul(sigmoid9_out1);
        let linear25_out1 = self.linear25.forward(mul34_out1);
        let mul35_out1 = linear25_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add34_out1 = layernormalization12_out1.add(mul35_out1);
        let layernormalization14_out1 = {
            let dtype = add34_out1.clone().dtype();
            self.layernormalization14
                .forward(add34_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat22_out1 = burn::tensor::Tensor::cat(
            [gather5_out1.clone(), layernormalization14_out1.clone()].into(),
            1,
        );
        let transpose40_out1 = gather5_out1.permute([1, 0, 2]);
        let slice15_out1 = transpose40_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose41_out1 = slice15_out1.permute([1, 0, 2]);
        let concat23_out1 = burn::tensor::Tensor::cat(
            [transpose41_out1, layernormalization14_out1.clone()].into(),
            1,
        );
        let linear26_out1 = self.linear26.forward(layernormalization14_out1);
        let reshape16_out1 = linear26_out1.reshape([1, -1, 8, 128]);
        let linear27_out1 = self.linear27.forward(concat22_out1.clone());
        let reshape17_out1 = linear27_out1.reshape([1, -1, 8, 128]);
        let linear28_out1 = self.linear28.forward(concat22_out1);
        let reshape18_out1 = linear28_out1.reshape([1, -1, 8, 128]);
        let transpose42_out1 = reshape17_out1.permute([0, 2, 1, 3]);
        let transpose43_out1 = reshape18_out1.permute([0, 2, 1, 3]);
        let linear29_out1 = self.linear29.forward(transpose6_out1);
        let reshape19_out1 = linear29_out1.reshape([1, -1, 8, 128]);
        let constant57_out1 = self.constant57.val();
        let add35_out1 = reshape16_out1
            .clone()
            .add((constant57_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose44_out1 = add35_out1.permute([0, 2, 1, 3]);
        let constant58_out1 = self.constant58.val();
        let add36_out1 = reshape16_out1
            .add((constant58_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose45_out1 = add36_out1.permute([0, 2, 1, 3]);
        let transpose46_out1 = reshape19_out1.permute([0, 2, 3, 1]);
        let matmul36_out1 = transpose45_out1.matmul(transpose46_out1);
        let pad5_out1 = matmul36_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape20_out1 = pad5_out1.reshape(concat16_out1);
        let slice16_out1 = reshape20_out1.slice(s![.., .., 1.., ..]);
        let reshape21_out1 = slice16_out1.reshape(concat17_out1);
        let transpose47_out1 = reshape21_out1.permute([3, 0, 1, 2]);
        let slice17_out1 = transpose47_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose48_out1 = slice17_out1.permute([1, 2, 3, 0]);
        let mul36_out1 = transpose48_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where7_out1 = mul36_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention3_out1,) = {
            let q = transpose44_out1;
            let k = transpose42_out1;
            let v = transpose43_out1;
            let score_bias = where7_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention3_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention3_out1 = {
                let dims = attention3_out1.dims();
                attention3_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention3_out1,)
        };
        let where8_out1 = attention3_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose49_out1 = where8_out1.permute([0, 2, 1, 3]);
        let reshape22_out1 = transpose49_out1.reshape([1, -1, 1024]);
        let linear30_out1 = self.linear30.forward(reshape22_out1);
        let add37_out1 = add34_out1.add(linear30_out1);
        (add37_out1, gather6_out1, concat23_out1, slice14_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule6 {
    layernormalization15: LayerNorm,
    squeeze6: burn::module::Param<Tensor<2>>,
    conv1d3: Conv1d,
    layernormalization16: LayerNorm,
    squeeze7: burn::module::Param<Tensor<2>>,
    layernormalization17: LayerNorm,
    linear31: Linear,
    linear32: Linear,
    layernormalization18: LayerNorm,
    layernormalization19: LayerNorm,
    linear33: Linear,
    linear34: Linear,
    layernormalization20: LayerNorm,
    linear35: Linear,
    linear36: Linear,
    linear37: Linear,
    linear38: Linear,
    constant74: burn::module::Param<Tensor<2>>,
    constant75: burn::module::Param<Tensor<2>>,
    linear39: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule6 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization15 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze6: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d3 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization16 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze7: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization17 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear31 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear32 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization18 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization19 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear33 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear34 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization20 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear35 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear36 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear37 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear38 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant74: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant75: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear39 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization15,
            squeeze6,
            conv1d3,
            layernormalization16,
            squeeze7,
            layernormalization17,
            linear31,
            linear32,
            layernormalization18,
            layernormalization19,
            linear33,
            linear34,
            layernormalization20,
            linear35,
            linear36,
            linear37,
            linear38,
            constant74,
            constant75,
            linear39,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add37_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather6_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization15_out1 = {
            let dtype = add37_out1.clone().dtype();
            self.layernormalization15
                .forward(add37_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze6_out1 = self.squeeze6.val();
        let transpose50_out1 = squeeze6_out1.permute([1, 0]);
        let matmul38_out1 = layernormalization15_out1
            .matmul(transpose50_out1.unsqueeze::<3usize>());
        let split_tensors = matmul38_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split3_out1, split3_out2] = split_tensors.try_into().unwrap();
        let sigmoid10_out1 = burn::tensor::activation::sigmoid(split3_out2);
        let mul37_out1 = split3_out1.mul(sigmoid10_out1);
        let transpose51_out1 = mul37_out1.permute([0, 2, 1]);
        let where9_out1 = transpose51_out1.mask_fill(unsqueeze33_out1, constant453_out1);
        let pad6_out1 = where9_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat24_out1 = burn::tensor::Tensor::cat(
            [gather6_out1, pad6_out1].into(),
            2,
        );
        let slice18_out1 = concat24_out1.clone().slice(s![.., .., - 8..]);
        let conv1d3_out1 = self.conv1d3.forward(concat24_out1);
        let transpose52_out1 = conv1d3_out1.permute([0, 2, 1]);
        let layernormalization16_out1 = {
            let dtype = transpose52_out1.dtype();
            self.layernormalization16
                .forward(transpose52_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid11_out1 = burn::tensor::activation::sigmoid(
            layernormalization16_out1.clone(),
        );
        let mul38_out1 = layernormalization16_out1.mul(sigmoid11_out1);
        let squeeze7_out1 = self.squeeze7.val();
        let transpose53_out1 = squeeze7_out1.permute([1, 0]);
        let matmul39_out1 = mul38_out1.matmul(transpose53_out1.unsqueeze::<3usize>());
        let add38_out1 = add37_out1.add(matmul39_out1);
        let layernormalization17_out1 = {
            let dtype = add38_out1.clone().dtype();
            self.layernormalization17
                .forward(add38_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear31_out1 = self.linear31.forward(layernormalization17_out1);
        let sigmoid12_out1 = burn::tensor::activation::sigmoid(linear31_out1.clone());
        let mul39_out1 = linear31_out1.mul(sigmoid12_out1);
        let linear32_out1 = self.linear32.forward(mul39_out1);
        let mul40_out1 = linear32_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add39_out1 = add38_out1.add(mul40_out1);
        let layernormalization18_out1 = {
            let dtype = add39_out1.dtype();
            self.layernormalization18
                .forward(add39_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather7_out1 = {
            let sliced = transpose1_out1.slice(s![3, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather8_out1 = {
            let sliced = transpose2_out1.slice(s![3, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization19_out1 = {
            let dtype = layernormalization18_out1.clone().dtype();
            self.layernormalization19
                .forward(
                    layernormalization18_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear33_out1 = self.linear33.forward(layernormalization19_out1);
        let sigmoid13_out1 = burn::tensor::activation::sigmoid(linear33_out1.clone());
        let mul41_out1 = linear33_out1.mul(sigmoid13_out1);
        let linear34_out1 = self.linear34.forward(mul41_out1);
        let mul42_out1 = linear34_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add40_out1 = layernormalization18_out1.add(mul42_out1);
        let layernormalization20_out1 = {
            let dtype = add40_out1.clone().dtype();
            self.layernormalization20
                .forward(add40_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat25_out1 = burn::tensor::Tensor::cat(
            [gather7_out1.clone(), layernormalization20_out1.clone()].into(),
            1,
        );
        let transpose54_out1 = gather7_out1.permute([1, 0, 2]);
        let slice19_out1 = transpose54_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose55_out1 = slice19_out1.permute([1, 0, 2]);
        let concat26_out1 = burn::tensor::Tensor::cat(
            [transpose55_out1, layernormalization20_out1.clone()].into(),
            1,
        );
        let linear35_out1 = self.linear35.forward(layernormalization20_out1);
        let reshape23_out1 = linear35_out1.reshape([1, -1, 8, 128]);
        let linear36_out1 = self.linear36.forward(concat25_out1.clone());
        let reshape24_out1 = linear36_out1.reshape([1, -1, 8, 128]);
        let linear37_out1 = self.linear37.forward(concat25_out1);
        let reshape25_out1 = linear37_out1.reshape([1, -1, 8, 128]);
        let transpose56_out1 = reshape24_out1.permute([0, 2, 1, 3]);
        let transpose57_out1 = reshape25_out1.permute([0, 2, 1, 3]);
        let linear38_out1 = self.linear38.forward(transpose6_out1);
        let reshape26_out1 = linear38_out1.reshape([1, -1, 8, 128]);
        let constant74_out1 = self.constant74.val();
        let add41_out1 = reshape23_out1
            .clone()
            .add((constant74_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose58_out1 = add41_out1.permute([0, 2, 1, 3]);
        let constant75_out1 = self.constant75.val();
        let add42_out1 = reshape23_out1
            .add((constant75_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose59_out1 = add42_out1.permute([0, 2, 1, 3]);
        let transpose60_out1 = reshape26_out1.permute([0, 2, 3, 1]);
        let matmul48_out1 = transpose59_out1.matmul(transpose60_out1);
        let pad7_out1 = matmul48_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape27_out1 = pad7_out1.reshape(concat16_out1);
        let slice20_out1 = reshape27_out1.slice(s![.., .., 1.., ..]);
        let reshape28_out1 = slice20_out1.reshape(concat17_out1);
        let transpose61_out1 = reshape28_out1.permute([3, 0, 1, 2]);
        let slice21_out1 = transpose61_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose62_out1 = slice21_out1.permute([1, 2, 3, 0]);
        let mul43_out1 = transpose62_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where10_out1 = mul43_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention4_out1,) = {
            let q = transpose58_out1;
            let k = transpose56_out1;
            let v = transpose57_out1;
            let score_bias = where10_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention4_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention4_out1 = {
                let dims = attention4_out1.dims();
                attention4_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention4_out1,)
        };
        let where11_out1 = attention4_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose63_out1 = where11_out1.permute([0, 2, 1, 3]);
        let reshape29_out1 = transpose63_out1.reshape([1, -1, 1024]);
        let linear39_out1 = self.linear39.forward(reshape29_out1);
        let add43_out1 = add40_out1.add(linear39_out1);
        (add43_out1, gather8_out1, concat26_out1, slice18_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule7 {
    layernormalization21: LayerNorm,
    squeeze8: burn::module::Param<Tensor<2>>,
    conv1d4: Conv1d,
    layernormalization22: LayerNorm,
    squeeze9: burn::module::Param<Tensor<2>>,
    layernormalization23: LayerNorm,
    linear40: Linear,
    linear41: Linear,
    layernormalization24: LayerNorm,
    layernormalization25: LayerNorm,
    linear42: Linear,
    linear43: Linear,
    layernormalization26: LayerNorm,
    linear44: Linear,
    linear45: Linear,
    linear46: Linear,
    linear47: Linear,
    constant91: burn::module::Param<Tensor<2>>,
    constant92: burn::module::Param<Tensor<2>>,
    linear48: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule7 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization21 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze8: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d4 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization22 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze9: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization23 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear40 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear41 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization24 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization25 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear42 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear43 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization26 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear44 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear45 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear46 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear47 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant91: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant92: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear48 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization21,
            squeeze8,
            conv1d4,
            layernormalization22,
            squeeze9,
            layernormalization23,
            linear40,
            linear41,
            layernormalization24,
            layernormalization25,
            linear42,
            linear43,
            layernormalization26,
            linear44,
            linear45,
            linear46,
            linear47,
            constant91,
            constant92,
            linear48,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add43_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather8_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization21_out1 = {
            let dtype = add43_out1.clone().dtype();
            self.layernormalization21
                .forward(add43_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze8_out1 = self.squeeze8.val();
        let transpose64_out1 = squeeze8_out1.permute([1, 0]);
        let matmul50_out1 = layernormalization21_out1
            .matmul(transpose64_out1.unsqueeze::<3usize>());
        let split_tensors = matmul50_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split4_out1, split4_out2] = split_tensors.try_into().unwrap();
        let sigmoid14_out1 = burn::tensor::activation::sigmoid(split4_out2);
        let mul44_out1 = split4_out1.mul(sigmoid14_out1);
        let transpose65_out1 = mul44_out1.permute([0, 2, 1]);
        let where12_out1 = transpose65_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad8_out1 = where12_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat27_out1 = burn::tensor::Tensor::cat(
            [gather8_out1, pad8_out1].into(),
            2,
        );
        let slice22_out1 = concat27_out1.clone().slice(s![.., .., - 8..]);
        let conv1d4_out1 = self.conv1d4.forward(concat27_out1);
        let transpose66_out1 = conv1d4_out1.permute([0, 2, 1]);
        let layernormalization22_out1 = {
            let dtype = transpose66_out1.dtype();
            self.layernormalization22
                .forward(transpose66_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid15_out1 = burn::tensor::activation::sigmoid(
            layernormalization22_out1.clone(),
        );
        let mul45_out1 = layernormalization22_out1.mul(sigmoid15_out1);
        let squeeze9_out1 = self.squeeze9.val();
        let transpose67_out1 = squeeze9_out1.permute([1, 0]);
        let matmul51_out1 = mul45_out1.matmul(transpose67_out1.unsqueeze::<3usize>());
        let add44_out1 = add43_out1.add(matmul51_out1);
        let layernormalization23_out1 = {
            let dtype = add44_out1.clone().dtype();
            self.layernormalization23
                .forward(add44_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear40_out1 = self.linear40.forward(layernormalization23_out1);
        let sigmoid16_out1 = burn::tensor::activation::sigmoid(linear40_out1.clone());
        let mul46_out1 = linear40_out1.mul(sigmoid16_out1);
        let linear41_out1 = self.linear41.forward(mul46_out1);
        let mul47_out1 = linear41_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add45_out1 = add44_out1.add(mul47_out1);
        let layernormalization24_out1 = {
            let dtype = add45_out1.dtype();
            self.layernormalization24
                .forward(add45_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather9_out1 = {
            let sliced = transpose1_out1.slice(s![4, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather10_out1 = {
            let sliced = transpose2_out1.slice(s![4, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization25_out1 = {
            let dtype = layernormalization24_out1.clone().dtype();
            self.layernormalization25
                .forward(
                    layernormalization24_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear42_out1 = self.linear42.forward(layernormalization25_out1);
        let sigmoid17_out1 = burn::tensor::activation::sigmoid(linear42_out1.clone());
        let mul48_out1 = linear42_out1.mul(sigmoid17_out1);
        let linear43_out1 = self.linear43.forward(mul48_out1);
        let mul49_out1 = linear43_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add46_out1 = layernormalization24_out1.add(mul49_out1);
        let layernormalization26_out1 = {
            let dtype = add46_out1.clone().dtype();
            self.layernormalization26
                .forward(add46_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat28_out1 = burn::tensor::Tensor::cat(
            [gather9_out1.clone(), layernormalization26_out1.clone()].into(),
            1,
        );
        let transpose68_out1 = gather9_out1.permute([1, 0, 2]);
        let slice23_out1 = transpose68_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose69_out1 = slice23_out1.permute([1, 0, 2]);
        let concat29_out1 = burn::tensor::Tensor::cat(
            [transpose69_out1, layernormalization26_out1.clone()].into(),
            1,
        );
        let linear44_out1 = self.linear44.forward(layernormalization26_out1);
        let reshape30_out1 = linear44_out1.reshape([1, -1, 8, 128]);
        let linear45_out1 = self.linear45.forward(concat28_out1.clone());
        let reshape31_out1 = linear45_out1.reshape([1, -1, 8, 128]);
        let linear46_out1 = self.linear46.forward(concat28_out1);
        let reshape32_out1 = linear46_out1.reshape([1, -1, 8, 128]);
        let transpose70_out1 = reshape31_out1.permute([0, 2, 1, 3]);
        let transpose71_out1 = reshape32_out1.permute([0, 2, 1, 3]);
        let linear47_out1 = self.linear47.forward(transpose6_out1);
        let reshape33_out1 = linear47_out1.reshape([1, -1, 8, 128]);
        let constant91_out1 = self.constant91.val();
        let add47_out1 = reshape30_out1
            .clone()
            .add((constant91_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose72_out1 = add47_out1.permute([0, 2, 1, 3]);
        let constant92_out1 = self.constant92.val();
        let add48_out1 = reshape30_out1
            .add((constant92_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose73_out1 = add48_out1.permute([0, 2, 1, 3]);
        let transpose74_out1 = reshape33_out1.permute([0, 2, 3, 1]);
        let matmul60_out1 = transpose73_out1.matmul(transpose74_out1);
        let pad9_out1 = matmul60_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape34_out1 = pad9_out1.reshape(concat16_out1);
        let slice24_out1 = reshape34_out1.slice(s![.., .., 1.., ..]);
        let reshape35_out1 = slice24_out1.reshape(concat17_out1);
        let transpose75_out1 = reshape35_out1.permute([3, 0, 1, 2]);
        let slice25_out1 = transpose75_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose76_out1 = slice25_out1.permute([1, 2, 3, 0]);
        let mul50_out1 = transpose76_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where13_out1 = mul50_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention5_out1,) = {
            let q = transpose72_out1;
            let k = transpose70_out1;
            let v = transpose71_out1;
            let score_bias = where13_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention5_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention5_out1 = {
                let dims = attention5_out1.dims();
                attention5_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention5_out1,)
        };
        let where14_out1 = attention5_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose77_out1 = where14_out1.permute([0, 2, 1, 3]);
        let reshape36_out1 = transpose77_out1.reshape([1, -1, 1024]);
        let linear48_out1 = self.linear48.forward(reshape36_out1);
        let add49_out1 = add46_out1.add(linear48_out1);
        (add49_out1, gather10_out1, concat29_out1, slice22_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule8 {
    layernormalization27: LayerNorm,
    squeeze10: burn::module::Param<Tensor<2>>,
    conv1d5: Conv1d,
    layernormalization28: LayerNorm,
    squeeze11: burn::module::Param<Tensor<2>>,
    layernormalization29: LayerNorm,
    linear49: Linear,
    linear50: Linear,
    layernormalization30: LayerNorm,
    layernormalization31: LayerNorm,
    linear51: Linear,
    linear52: Linear,
    layernormalization32: LayerNorm,
    linear53: Linear,
    linear54: Linear,
    linear55: Linear,
    linear56: Linear,
    constant108: burn::module::Param<Tensor<2>>,
    constant109: burn::module::Param<Tensor<2>>,
    linear57: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule8 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization27 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze10: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d5 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization28 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze11: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization29 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear49 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear50 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization30 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization31 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear51 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear52 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization32 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear53 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear54 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear55 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear56 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant108: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant109: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear57 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization27,
            squeeze10,
            conv1d5,
            layernormalization28,
            squeeze11,
            layernormalization29,
            linear49,
            linear50,
            layernormalization30,
            layernormalization31,
            linear51,
            linear52,
            layernormalization32,
            linear53,
            linear54,
            linear55,
            linear56,
            constant108,
            constant109,
            linear57,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add49_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather10_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization27_out1 = {
            let dtype = add49_out1.clone().dtype();
            self.layernormalization27
                .forward(add49_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze10_out1 = self.squeeze10.val();
        let transpose78_out1 = squeeze10_out1.permute([1, 0]);
        let matmul62_out1 = layernormalization27_out1
            .matmul(transpose78_out1.unsqueeze::<3usize>());
        let split_tensors = matmul62_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split5_out1, split5_out2] = split_tensors.try_into().unwrap();
        let sigmoid18_out1 = burn::tensor::activation::sigmoid(split5_out2);
        let mul51_out1 = split5_out1.mul(sigmoid18_out1);
        let transpose79_out1 = mul51_out1.permute([0, 2, 1]);
        let where15_out1 = transpose79_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad10_out1 = where15_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat30_out1 = burn::tensor::Tensor::cat(
            [gather10_out1, pad10_out1].into(),
            2,
        );
        let slice26_out1 = concat30_out1.clone().slice(s![.., .., - 8..]);
        let conv1d5_out1 = self.conv1d5.forward(concat30_out1);
        let transpose80_out1 = conv1d5_out1.permute([0, 2, 1]);
        let layernormalization28_out1 = {
            let dtype = transpose80_out1.dtype();
            self.layernormalization28
                .forward(transpose80_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid19_out1 = burn::tensor::activation::sigmoid(
            layernormalization28_out1.clone(),
        );
        let mul52_out1 = layernormalization28_out1.mul(sigmoid19_out1);
        let squeeze11_out1 = self.squeeze11.val();
        let transpose81_out1 = squeeze11_out1.permute([1, 0]);
        let matmul63_out1 = mul52_out1.matmul(transpose81_out1.unsqueeze::<3usize>());
        let add50_out1 = add49_out1.add(matmul63_out1);
        let layernormalization29_out1 = {
            let dtype = add50_out1.clone().dtype();
            self.layernormalization29
                .forward(add50_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear49_out1 = self.linear49.forward(layernormalization29_out1);
        let sigmoid20_out1 = burn::tensor::activation::sigmoid(linear49_out1.clone());
        let mul53_out1 = linear49_out1.mul(sigmoid20_out1);
        let linear50_out1 = self.linear50.forward(mul53_out1);
        let mul54_out1 = linear50_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add51_out1 = add50_out1.add(mul54_out1);
        let layernormalization30_out1 = {
            let dtype = add51_out1.dtype();
            self.layernormalization30
                .forward(add51_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather11_out1 = {
            let sliced = transpose1_out1.slice(s![5, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather12_out1 = {
            let sliced = transpose2_out1.slice(s![5, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization31_out1 = {
            let dtype = layernormalization30_out1.clone().dtype();
            self.layernormalization31
                .forward(
                    layernormalization30_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear51_out1 = self.linear51.forward(layernormalization31_out1);
        let sigmoid21_out1 = burn::tensor::activation::sigmoid(linear51_out1.clone());
        let mul55_out1 = linear51_out1.mul(sigmoid21_out1);
        let linear52_out1 = self.linear52.forward(mul55_out1);
        let mul56_out1 = linear52_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add52_out1 = layernormalization30_out1.add(mul56_out1);
        let layernormalization32_out1 = {
            let dtype = add52_out1.clone().dtype();
            self.layernormalization32
                .forward(add52_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat31_out1 = burn::tensor::Tensor::cat(
            [gather11_out1.clone(), layernormalization32_out1.clone()].into(),
            1,
        );
        let transpose82_out1 = gather11_out1.permute([1, 0, 2]);
        let slice27_out1 = transpose82_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose83_out1 = slice27_out1.permute([1, 0, 2]);
        let concat32_out1 = burn::tensor::Tensor::cat(
            [transpose83_out1, layernormalization32_out1.clone()].into(),
            1,
        );
        let linear53_out1 = self.linear53.forward(layernormalization32_out1);
        let reshape37_out1 = linear53_out1.reshape([1, -1, 8, 128]);
        let linear54_out1 = self.linear54.forward(concat31_out1.clone());
        let reshape38_out1 = linear54_out1.reshape([1, -1, 8, 128]);
        let linear55_out1 = self.linear55.forward(concat31_out1);
        let reshape39_out1 = linear55_out1.reshape([1, -1, 8, 128]);
        let transpose84_out1 = reshape38_out1.permute([0, 2, 1, 3]);
        let transpose85_out1 = reshape39_out1.permute([0, 2, 1, 3]);
        let linear56_out1 = self.linear56.forward(transpose6_out1);
        let reshape40_out1 = linear56_out1.reshape([1, -1, 8, 128]);
        let constant108_out1 = self.constant108.val();
        let add53_out1 = reshape37_out1
            .clone()
            .add((constant108_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose86_out1 = add53_out1.permute([0, 2, 1, 3]);
        let constant109_out1 = self.constant109.val();
        let add54_out1 = reshape37_out1
            .add((constant109_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose87_out1 = add54_out1.permute([0, 2, 1, 3]);
        let transpose88_out1 = reshape40_out1.permute([0, 2, 3, 1]);
        let matmul72_out1 = transpose87_out1.matmul(transpose88_out1);
        let pad11_out1 = matmul72_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape41_out1 = pad11_out1.reshape(concat16_out1);
        let slice28_out1 = reshape41_out1.slice(s![.., .., 1.., ..]);
        let reshape42_out1 = slice28_out1.reshape(concat17_out1);
        let transpose89_out1 = reshape42_out1.permute([3, 0, 1, 2]);
        let slice29_out1 = transpose89_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose90_out1 = slice29_out1.permute([1, 2, 3, 0]);
        let mul57_out1 = transpose90_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where16_out1 = mul57_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention6_out1,) = {
            let q = transpose86_out1;
            let k = transpose84_out1;
            let v = transpose85_out1;
            let score_bias = where16_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention6_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention6_out1 = {
                let dims = attention6_out1.dims();
                attention6_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention6_out1,)
        };
        let where17_out1 = attention6_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose91_out1 = where17_out1.permute([0, 2, 1, 3]);
        let reshape43_out1 = transpose91_out1.reshape([1, -1, 1024]);
        let linear57_out1 = self.linear57.forward(reshape43_out1);
        let add55_out1 = add52_out1.add(linear57_out1);
        (add55_out1, gather12_out1, concat32_out1, slice26_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule9 {
    layernormalization33: LayerNorm,
    squeeze12: burn::module::Param<Tensor<2>>,
    conv1d6: Conv1d,
    layernormalization34: LayerNorm,
    squeeze13: burn::module::Param<Tensor<2>>,
    layernormalization35: LayerNorm,
    linear58: Linear,
    linear59: Linear,
    layernormalization36: LayerNorm,
    layernormalization37: LayerNorm,
    linear60: Linear,
    linear61: Linear,
    layernormalization38: LayerNorm,
    linear62: Linear,
    linear63: Linear,
    linear64: Linear,
    linear65: Linear,
    constant125: burn::module::Param<Tensor<2>>,
    constant126: burn::module::Param<Tensor<2>>,
    linear66: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule9 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization33 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze12: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d6 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization34 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze13: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization35 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear58 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear59 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization36 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization37 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear60 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear61 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization38 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear62 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear63 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear64 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear65 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant125: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant126: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear66 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization33,
            squeeze12,
            conv1d6,
            layernormalization34,
            squeeze13,
            layernormalization35,
            linear58,
            linear59,
            layernormalization36,
            layernormalization37,
            linear60,
            linear61,
            layernormalization38,
            linear62,
            linear63,
            linear64,
            linear65,
            constant125,
            constant126,
            linear66,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add55_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather12_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization33_out1 = {
            let dtype = add55_out1.clone().dtype();
            self.layernormalization33
                .forward(add55_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze12_out1 = self.squeeze12.val();
        let transpose92_out1 = squeeze12_out1.permute([1, 0]);
        let matmul74_out1 = layernormalization33_out1
            .matmul(transpose92_out1.unsqueeze::<3usize>());
        let split_tensors = matmul74_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split6_out1, split6_out2] = split_tensors.try_into().unwrap();
        let sigmoid22_out1 = burn::tensor::activation::sigmoid(split6_out2);
        let mul58_out1 = split6_out1.mul(sigmoid22_out1);
        let transpose93_out1 = mul58_out1.permute([0, 2, 1]);
        let where18_out1 = transpose93_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad12_out1 = where18_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat33_out1 = burn::tensor::Tensor::cat(
            [gather12_out1, pad12_out1].into(),
            2,
        );
        let slice30_out1 = concat33_out1.clone().slice(s![.., .., - 8..]);
        let conv1d6_out1 = self.conv1d6.forward(concat33_out1);
        let transpose94_out1 = conv1d6_out1.permute([0, 2, 1]);
        let layernormalization34_out1 = {
            let dtype = transpose94_out1.dtype();
            self.layernormalization34
                .forward(transpose94_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid23_out1 = burn::tensor::activation::sigmoid(
            layernormalization34_out1.clone(),
        );
        let mul59_out1 = layernormalization34_out1.mul(sigmoid23_out1);
        let squeeze13_out1 = self.squeeze13.val();
        let transpose95_out1 = squeeze13_out1.permute([1, 0]);
        let matmul75_out1 = mul59_out1.matmul(transpose95_out1.unsqueeze::<3usize>());
        let add56_out1 = add55_out1.add(matmul75_out1);
        let layernormalization35_out1 = {
            let dtype = add56_out1.clone().dtype();
            self.layernormalization35
                .forward(add56_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear58_out1 = self.linear58.forward(layernormalization35_out1);
        let sigmoid24_out1 = burn::tensor::activation::sigmoid(linear58_out1.clone());
        let mul60_out1 = linear58_out1.mul(sigmoid24_out1);
        let linear59_out1 = self.linear59.forward(mul60_out1);
        let mul61_out1 = linear59_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add57_out1 = add56_out1.add(mul61_out1);
        let layernormalization36_out1 = {
            let dtype = add57_out1.dtype();
            self.layernormalization36
                .forward(add57_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather13_out1 = {
            let sliced = transpose1_out1.slice(s![6, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather14_out1 = {
            let sliced = transpose2_out1.slice(s![6, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization37_out1 = {
            let dtype = layernormalization36_out1.clone().dtype();
            self.layernormalization37
                .forward(
                    layernormalization36_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear60_out1 = self.linear60.forward(layernormalization37_out1);
        let sigmoid25_out1 = burn::tensor::activation::sigmoid(linear60_out1.clone());
        let mul62_out1 = linear60_out1.mul(sigmoid25_out1);
        let linear61_out1 = self.linear61.forward(mul62_out1);
        let mul63_out1 = linear61_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add58_out1 = layernormalization36_out1.add(mul63_out1);
        let layernormalization38_out1 = {
            let dtype = add58_out1.clone().dtype();
            self.layernormalization38
                .forward(add58_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat34_out1 = burn::tensor::Tensor::cat(
            [gather13_out1.clone(), layernormalization38_out1.clone()].into(),
            1,
        );
        let transpose96_out1 = gather13_out1.permute([1, 0, 2]);
        let slice31_out1 = transpose96_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose97_out1 = slice31_out1.permute([1, 0, 2]);
        let concat35_out1 = burn::tensor::Tensor::cat(
            [transpose97_out1, layernormalization38_out1.clone()].into(),
            1,
        );
        let linear62_out1 = self.linear62.forward(layernormalization38_out1);
        let reshape44_out1 = linear62_out1.reshape([1, -1, 8, 128]);
        let linear63_out1 = self.linear63.forward(concat34_out1.clone());
        let reshape45_out1 = linear63_out1.reshape([1, -1, 8, 128]);
        let linear64_out1 = self.linear64.forward(concat34_out1);
        let reshape46_out1 = linear64_out1.reshape([1, -1, 8, 128]);
        let transpose98_out1 = reshape45_out1.permute([0, 2, 1, 3]);
        let transpose99_out1 = reshape46_out1.permute([0, 2, 1, 3]);
        let linear65_out1 = self.linear65.forward(transpose6_out1);
        let reshape47_out1 = linear65_out1.reshape([1, -1, 8, 128]);
        let constant125_out1 = self.constant125.val();
        let add59_out1 = reshape44_out1
            .clone()
            .add((constant125_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose100_out1 = add59_out1.permute([0, 2, 1, 3]);
        let constant126_out1 = self.constant126.val();
        let add60_out1 = reshape44_out1
            .add((constant126_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose101_out1 = add60_out1.permute([0, 2, 1, 3]);
        let transpose102_out1 = reshape47_out1.permute([0, 2, 3, 1]);
        let matmul84_out1 = transpose101_out1.matmul(transpose102_out1);
        let pad13_out1 = matmul84_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape48_out1 = pad13_out1.reshape(concat16_out1);
        let slice32_out1 = reshape48_out1.slice(s![.., .., 1.., ..]);
        let reshape49_out1 = slice32_out1.reshape(concat17_out1);
        let transpose103_out1 = reshape49_out1.permute([3, 0, 1, 2]);
        let slice33_out1 = transpose103_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose104_out1 = slice33_out1.permute([1, 2, 3, 0]);
        let mul64_out1 = transpose104_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where19_out1 = mul64_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention7_out1,) = {
            let q = transpose100_out1;
            let k = transpose98_out1;
            let v = transpose99_out1;
            let score_bias = where19_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention7_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention7_out1 = {
                let dims = attention7_out1.dims();
                attention7_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention7_out1,)
        };
        let where20_out1 = attention7_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose105_out1 = where20_out1.permute([0, 2, 1, 3]);
        let reshape50_out1 = transpose105_out1.reshape([1, -1, 1024]);
        let linear66_out1 = self.linear66.forward(reshape50_out1);
        let add61_out1 = add58_out1.add(linear66_out1);
        (add61_out1, gather14_out1, concat35_out1, slice30_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule10 {
    layernormalization39: LayerNorm,
    squeeze14: burn::module::Param<Tensor<2>>,
    conv1d7: Conv1d,
    layernormalization40: LayerNorm,
    squeeze15: burn::module::Param<Tensor<2>>,
    layernormalization41: LayerNorm,
    linear67: Linear,
    linear68: Linear,
    layernormalization42: LayerNorm,
    layernormalization43: LayerNorm,
    linear69: Linear,
    linear70: Linear,
    layernormalization44: LayerNorm,
    linear71: Linear,
    linear72: Linear,
    linear73: Linear,
    linear74: Linear,
    constant142: burn::module::Param<Tensor<2>>,
    constant143: burn::module::Param<Tensor<2>>,
    linear75: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule10 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization39 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze14: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d7 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization40 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze15: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization41 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear67 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear68 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization42 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization43 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear69 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear70 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization44 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear71 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear72 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear73 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear74 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant142: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant143: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear75 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization39,
            squeeze14,
            conv1d7,
            layernormalization40,
            squeeze15,
            layernormalization41,
            linear67,
            linear68,
            layernormalization42,
            layernormalization43,
            linear69,
            linear70,
            layernormalization44,
            linear71,
            linear72,
            linear73,
            linear74,
            constant142,
            constant143,
            linear75,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add61_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather14_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization39_out1 = {
            let dtype = add61_out1.clone().dtype();
            self.layernormalization39
                .forward(add61_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze14_out1 = self.squeeze14.val();
        let transpose106_out1 = squeeze14_out1.permute([1, 0]);
        let matmul86_out1 = layernormalization39_out1
            .matmul(transpose106_out1.unsqueeze::<3usize>());
        let split_tensors = matmul86_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split7_out1, split7_out2] = split_tensors.try_into().unwrap();
        let sigmoid26_out1 = burn::tensor::activation::sigmoid(split7_out2);
        let mul65_out1 = split7_out1.mul(sigmoid26_out1);
        let transpose107_out1 = mul65_out1.permute([0, 2, 1]);
        let where21_out1 = transpose107_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad14_out1 = where21_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat36_out1 = burn::tensor::Tensor::cat(
            [gather14_out1, pad14_out1].into(),
            2,
        );
        let slice34_out1 = concat36_out1.clone().slice(s![.., .., - 8..]);
        let conv1d7_out1 = self.conv1d7.forward(concat36_out1);
        let transpose108_out1 = conv1d7_out1.permute([0, 2, 1]);
        let layernormalization40_out1 = {
            let dtype = transpose108_out1.dtype();
            self.layernormalization40
                .forward(transpose108_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid27_out1 = burn::tensor::activation::sigmoid(
            layernormalization40_out1.clone(),
        );
        let mul66_out1 = layernormalization40_out1.mul(sigmoid27_out1);
        let squeeze15_out1 = self.squeeze15.val();
        let transpose109_out1 = squeeze15_out1.permute([1, 0]);
        let matmul87_out1 = mul66_out1.matmul(transpose109_out1.unsqueeze::<3usize>());
        let add62_out1 = add61_out1.add(matmul87_out1);
        let layernormalization41_out1 = {
            let dtype = add62_out1.clone().dtype();
            self.layernormalization41
                .forward(add62_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear67_out1 = self.linear67.forward(layernormalization41_out1);
        let sigmoid28_out1 = burn::tensor::activation::sigmoid(linear67_out1.clone());
        let mul67_out1 = linear67_out1.mul(sigmoid28_out1);
        let linear68_out1 = self.linear68.forward(mul67_out1);
        let mul68_out1 = linear68_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add63_out1 = add62_out1.add(mul68_out1);
        let layernormalization42_out1 = {
            let dtype = add63_out1.dtype();
            self.layernormalization42
                .forward(add63_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather15_out1 = {
            let sliced = transpose1_out1.slice(s![7, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather16_out1 = {
            let sliced = transpose2_out1.slice(s![7, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization43_out1 = {
            let dtype = layernormalization42_out1.clone().dtype();
            self.layernormalization43
                .forward(
                    layernormalization42_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear69_out1 = self.linear69.forward(layernormalization43_out1);
        let sigmoid29_out1 = burn::tensor::activation::sigmoid(linear69_out1.clone());
        let mul69_out1 = linear69_out1.mul(sigmoid29_out1);
        let linear70_out1 = self.linear70.forward(mul69_out1);
        let mul70_out1 = linear70_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add64_out1 = layernormalization42_out1.add(mul70_out1);
        let layernormalization44_out1 = {
            let dtype = add64_out1.clone().dtype();
            self.layernormalization44
                .forward(add64_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat37_out1 = burn::tensor::Tensor::cat(
            [gather15_out1.clone(), layernormalization44_out1.clone()].into(),
            1,
        );
        let transpose110_out1 = gather15_out1.permute([1, 0, 2]);
        let slice35_out1 = transpose110_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose111_out1 = slice35_out1.permute([1, 0, 2]);
        let concat38_out1 = burn::tensor::Tensor::cat(
            [transpose111_out1, layernormalization44_out1.clone()].into(),
            1,
        );
        let linear71_out1 = self.linear71.forward(layernormalization44_out1);
        let reshape51_out1 = linear71_out1.reshape([1, -1, 8, 128]);
        let linear72_out1 = self.linear72.forward(concat37_out1.clone());
        let reshape52_out1 = linear72_out1.reshape([1, -1, 8, 128]);
        let linear73_out1 = self.linear73.forward(concat37_out1);
        let reshape53_out1 = linear73_out1.reshape([1, -1, 8, 128]);
        let transpose112_out1 = reshape52_out1.permute([0, 2, 1, 3]);
        let transpose113_out1 = reshape53_out1.permute([0, 2, 1, 3]);
        let linear74_out1 = self.linear74.forward(transpose6_out1);
        let reshape54_out1 = linear74_out1.reshape([1, -1, 8, 128]);
        let constant142_out1 = self.constant142.val();
        let add65_out1 = reshape51_out1
            .clone()
            .add((constant142_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose114_out1 = add65_out1.permute([0, 2, 1, 3]);
        let constant143_out1 = self.constant143.val();
        let add66_out1 = reshape51_out1
            .add((constant143_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose115_out1 = add66_out1.permute([0, 2, 1, 3]);
        let transpose116_out1 = reshape54_out1.permute([0, 2, 3, 1]);
        let matmul96_out1 = transpose115_out1.matmul(transpose116_out1);
        let pad15_out1 = matmul96_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape55_out1 = pad15_out1.reshape(concat16_out1);
        let slice36_out1 = reshape55_out1.slice(s![.., .., 1.., ..]);
        let reshape56_out1 = slice36_out1.reshape(concat17_out1);
        let transpose117_out1 = reshape56_out1.permute([3, 0, 1, 2]);
        let slice37_out1 = transpose117_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose118_out1 = slice37_out1.permute([1, 2, 3, 0]);
        let mul71_out1 = transpose118_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where22_out1 = mul71_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention8_out1,) = {
            let q = transpose114_out1;
            let k = transpose112_out1;
            let v = transpose113_out1;
            let score_bias = where22_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention8_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention8_out1 = {
                let dims = attention8_out1.dims();
                attention8_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention8_out1,)
        };
        let where23_out1 = attention8_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose119_out1 = where23_out1.permute([0, 2, 1, 3]);
        let reshape57_out1 = transpose119_out1.reshape([1, -1, 1024]);
        let linear75_out1 = self.linear75.forward(reshape57_out1);
        let add67_out1 = add64_out1.add(linear75_out1);
        (add67_out1, gather16_out1, concat38_out1, slice34_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule11 {
    layernormalization45: LayerNorm,
    squeeze16: burn::module::Param<Tensor<2>>,
    conv1d8: Conv1d,
    layernormalization46: LayerNorm,
    squeeze17: burn::module::Param<Tensor<2>>,
    layernormalization47: LayerNorm,
    linear76: Linear,
    linear77: Linear,
    layernormalization48: LayerNorm,
    layernormalization49: LayerNorm,
    linear78: Linear,
    linear79: Linear,
    layernormalization50: LayerNorm,
    linear80: Linear,
    linear81: Linear,
    linear82: Linear,
    linear83: Linear,
    constant159: burn::module::Param<Tensor<2>>,
    constant160: burn::module::Param<Tensor<2>>,
    linear84: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule11 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization45 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze16: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d8 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization46 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze17: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization47 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear76 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear77 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization48 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization49 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear78 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear79 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization50 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear80 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear81 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear82 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear83 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant159: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant160: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear84 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization45,
            squeeze16,
            conv1d8,
            layernormalization46,
            squeeze17,
            layernormalization47,
            linear76,
            linear77,
            layernormalization48,
            layernormalization49,
            linear78,
            linear79,
            layernormalization50,
            linear80,
            linear81,
            linear82,
            linear83,
            constant159,
            constant160,
            linear84,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add67_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather16_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization45_out1 = {
            let dtype = add67_out1.clone().dtype();
            self.layernormalization45
                .forward(add67_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze16_out1 = self.squeeze16.val();
        let transpose120_out1 = squeeze16_out1.permute([1, 0]);
        let matmul98_out1 = layernormalization45_out1
            .matmul(transpose120_out1.unsqueeze::<3usize>());
        let split_tensors = matmul98_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split8_out1, split8_out2] = split_tensors.try_into().unwrap();
        let sigmoid30_out1 = burn::tensor::activation::sigmoid(split8_out2);
        let mul72_out1 = split8_out1.mul(sigmoid30_out1);
        let transpose121_out1 = mul72_out1.permute([0, 2, 1]);
        let where24_out1 = transpose121_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad16_out1 = where24_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat39_out1 = burn::tensor::Tensor::cat(
            [gather16_out1, pad16_out1].into(),
            2,
        );
        let slice38_out1 = concat39_out1.clone().slice(s![.., .., - 8..]);
        let conv1d8_out1 = self.conv1d8.forward(concat39_out1);
        let transpose122_out1 = conv1d8_out1.permute([0, 2, 1]);
        let layernormalization46_out1 = {
            let dtype = transpose122_out1.dtype();
            self.layernormalization46
                .forward(transpose122_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid31_out1 = burn::tensor::activation::sigmoid(
            layernormalization46_out1.clone(),
        );
        let mul73_out1 = layernormalization46_out1.mul(sigmoid31_out1);
        let squeeze17_out1 = self.squeeze17.val();
        let transpose123_out1 = squeeze17_out1.permute([1, 0]);
        let matmul99_out1 = mul73_out1.matmul(transpose123_out1.unsqueeze::<3usize>());
        let add68_out1 = add67_out1.add(matmul99_out1);
        let layernormalization47_out1 = {
            let dtype = add68_out1.clone().dtype();
            self.layernormalization47
                .forward(add68_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear76_out1 = self.linear76.forward(layernormalization47_out1);
        let sigmoid32_out1 = burn::tensor::activation::sigmoid(linear76_out1.clone());
        let mul74_out1 = linear76_out1.mul(sigmoid32_out1);
        let linear77_out1 = self.linear77.forward(mul74_out1);
        let mul75_out1 = linear77_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add69_out1 = add68_out1.add(mul75_out1);
        let layernormalization48_out1 = {
            let dtype = add69_out1.dtype();
            self.layernormalization48
                .forward(add69_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather17_out1 = {
            let sliced = transpose1_out1.slice(s![8, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather18_out1 = {
            let sliced = transpose2_out1.slice(s![8, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization49_out1 = {
            let dtype = layernormalization48_out1.clone().dtype();
            self.layernormalization49
                .forward(
                    layernormalization48_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear78_out1 = self.linear78.forward(layernormalization49_out1);
        let sigmoid33_out1 = burn::tensor::activation::sigmoid(linear78_out1.clone());
        let mul76_out1 = linear78_out1.mul(sigmoid33_out1);
        let linear79_out1 = self.linear79.forward(mul76_out1);
        let mul77_out1 = linear79_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add70_out1 = layernormalization48_out1.add(mul77_out1);
        let layernormalization50_out1 = {
            let dtype = add70_out1.clone().dtype();
            self.layernormalization50
                .forward(add70_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat40_out1 = burn::tensor::Tensor::cat(
            [gather17_out1.clone(), layernormalization50_out1.clone()].into(),
            1,
        );
        let transpose124_out1 = gather17_out1.permute([1, 0, 2]);
        let slice39_out1 = transpose124_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose125_out1 = slice39_out1.permute([1, 0, 2]);
        let concat41_out1 = burn::tensor::Tensor::cat(
            [transpose125_out1, layernormalization50_out1.clone()].into(),
            1,
        );
        let linear80_out1 = self.linear80.forward(layernormalization50_out1);
        let reshape58_out1 = linear80_out1.reshape([1, -1, 8, 128]);
        let linear81_out1 = self.linear81.forward(concat40_out1.clone());
        let reshape59_out1 = linear81_out1.reshape([1, -1, 8, 128]);
        let linear82_out1 = self.linear82.forward(concat40_out1);
        let reshape60_out1 = linear82_out1.reshape([1, -1, 8, 128]);
        let transpose126_out1 = reshape59_out1.permute([0, 2, 1, 3]);
        let transpose127_out1 = reshape60_out1.permute([0, 2, 1, 3]);
        let linear83_out1 = self.linear83.forward(transpose6_out1);
        let reshape61_out1 = linear83_out1.reshape([1, -1, 8, 128]);
        let constant159_out1 = self.constant159.val();
        let add71_out1 = reshape58_out1
            .clone()
            .add((constant159_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose128_out1 = add71_out1.permute([0, 2, 1, 3]);
        let constant160_out1 = self.constant160.val();
        let add72_out1 = reshape58_out1
            .add((constant160_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose129_out1 = add72_out1.permute([0, 2, 1, 3]);
        let transpose130_out1 = reshape61_out1.permute([0, 2, 3, 1]);
        let matmul108_out1 = transpose129_out1.matmul(transpose130_out1);
        let pad17_out1 = matmul108_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape62_out1 = pad17_out1.reshape(concat16_out1);
        let slice40_out1 = reshape62_out1.slice(s![.., .., 1.., ..]);
        let reshape63_out1 = slice40_out1.reshape(concat17_out1);
        let transpose131_out1 = reshape63_out1.permute([3, 0, 1, 2]);
        let slice41_out1 = transpose131_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose132_out1 = slice41_out1.permute([1, 2, 3, 0]);
        let mul78_out1 = transpose132_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where25_out1 = mul78_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention9_out1,) = {
            let q = transpose128_out1;
            let k = transpose126_out1;
            let v = transpose127_out1;
            let score_bias = where25_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention9_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention9_out1 = {
                let dims = attention9_out1.dims();
                attention9_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention9_out1,)
        };
        let where26_out1 = attention9_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose133_out1 = where26_out1.permute([0, 2, 1, 3]);
        let reshape64_out1 = transpose133_out1.reshape([1, -1, 1024]);
        let linear84_out1 = self.linear84.forward(reshape64_out1);
        let add73_out1 = add70_out1.add(linear84_out1);
        (add73_out1, gather18_out1, concat41_out1, slice38_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule12 {
    layernormalization51: LayerNorm,
    squeeze18: burn::module::Param<Tensor<2>>,
    conv1d9: Conv1d,
    layernormalization52: LayerNorm,
    squeeze19: burn::module::Param<Tensor<2>>,
    layernormalization53: LayerNorm,
    linear85: Linear,
    linear86: Linear,
    layernormalization54: LayerNorm,
    layernormalization55: LayerNorm,
    linear87: Linear,
    linear88: Linear,
    layernormalization56: LayerNorm,
    linear89: Linear,
    linear90: Linear,
    linear91: Linear,
    linear92: Linear,
    constant176: burn::module::Param<Tensor<2>>,
    constant177: burn::module::Param<Tensor<2>>,
    linear93: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule12 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization51 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze18: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d9 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization52 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze19: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization53 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear85 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear86 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization54 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization55 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear87 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear88 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization56 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear89 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear90 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear91 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear92 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant176: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant177: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear93 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization51,
            squeeze18,
            conv1d9,
            layernormalization52,
            squeeze19,
            layernormalization53,
            linear85,
            linear86,
            layernormalization54,
            layernormalization55,
            linear87,
            linear88,
            layernormalization56,
            linear89,
            linear90,
            linear91,
            linear92,
            constant176,
            constant177,
            linear93,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add73_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather18_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization51_out1 = {
            let dtype = add73_out1.clone().dtype();
            self.layernormalization51
                .forward(add73_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze18_out1 = self.squeeze18.val();
        let transpose134_out1 = squeeze18_out1.permute([1, 0]);
        let matmul110_out1 = layernormalization51_out1
            .matmul(transpose134_out1.unsqueeze::<3usize>());
        let split_tensors = matmul110_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split9_out1, split9_out2] = split_tensors.try_into().unwrap();
        let sigmoid34_out1 = burn::tensor::activation::sigmoid(split9_out2);
        let mul79_out1 = split9_out1.mul(sigmoid34_out1);
        let transpose135_out1 = mul79_out1.permute([0, 2, 1]);
        let where27_out1 = transpose135_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad18_out1 = where27_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat42_out1 = burn::tensor::Tensor::cat(
            [gather18_out1, pad18_out1].into(),
            2,
        );
        let slice42_out1 = concat42_out1.clone().slice(s![.., .., - 8..]);
        let conv1d9_out1 = self.conv1d9.forward(concat42_out1);
        let transpose136_out1 = conv1d9_out1.permute([0, 2, 1]);
        let layernormalization52_out1 = {
            let dtype = transpose136_out1.dtype();
            self.layernormalization52
                .forward(transpose136_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid35_out1 = burn::tensor::activation::sigmoid(
            layernormalization52_out1.clone(),
        );
        let mul80_out1 = layernormalization52_out1.mul(sigmoid35_out1);
        let squeeze19_out1 = self.squeeze19.val();
        let transpose137_out1 = squeeze19_out1.permute([1, 0]);
        let matmul111_out1 = mul80_out1.matmul(transpose137_out1.unsqueeze::<3usize>());
        let add74_out1 = add73_out1.add(matmul111_out1);
        let layernormalization53_out1 = {
            let dtype = add74_out1.clone().dtype();
            self.layernormalization53
                .forward(add74_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear85_out1 = self.linear85.forward(layernormalization53_out1);
        let sigmoid36_out1 = burn::tensor::activation::sigmoid(linear85_out1.clone());
        let mul81_out1 = linear85_out1.mul(sigmoid36_out1);
        let linear86_out1 = self.linear86.forward(mul81_out1);
        let mul82_out1 = linear86_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add75_out1 = add74_out1.add(mul82_out1);
        let layernormalization54_out1 = {
            let dtype = add75_out1.dtype();
            self.layernormalization54
                .forward(add75_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather19_out1 = {
            let sliced = transpose1_out1.slice(s![9, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather20_out1 = {
            let sliced = transpose2_out1.slice(s![9, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization55_out1 = {
            let dtype = layernormalization54_out1.clone().dtype();
            self.layernormalization55
                .forward(
                    layernormalization54_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear87_out1 = self.linear87.forward(layernormalization55_out1);
        let sigmoid37_out1 = burn::tensor::activation::sigmoid(linear87_out1.clone());
        let mul83_out1 = linear87_out1.mul(sigmoid37_out1);
        let linear88_out1 = self.linear88.forward(mul83_out1);
        let mul84_out1 = linear88_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add76_out1 = layernormalization54_out1.add(mul84_out1);
        let layernormalization56_out1 = {
            let dtype = add76_out1.clone().dtype();
            self.layernormalization56
                .forward(add76_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat43_out1 = burn::tensor::Tensor::cat(
            [gather19_out1.clone(), layernormalization56_out1.clone()].into(),
            1,
        );
        let transpose138_out1 = gather19_out1.permute([1, 0, 2]);
        let slice43_out1 = transpose138_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose139_out1 = slice43_out1.permute([1, 0, 2]);
        let concat44_out1 = burn::tensor::Tensor::cat(
            [transpose139_out1, layernormalization56_out1.clone()].into(),
            1,
        );
        let linear89_out1 = self.linear89.forward(layernormalization56_out1);
        let reshape65_out1 = linear89_out1.reshape([1, -1, 8, 128]);
        let linear90_out1 = self.linear90.forward(concat43_out1.clone());
        let reshape66_out1 = linear90_out1.reshape([1, -1, 8, 128]);
        let linear91_out1 = self.linear91.forward(concat43_out1);
        let reshape67_out1 = linear91_out1.reshape([1, -1, 8, 128]);
        let transpose140_out1 = reshape66_out1.permute([0, 2, 1, 3]);
        let transpose141_out1 = reshape67_out1.permute([0, 2, 1, 3]);
        let linear92_out1 = self.linear92.forward(transpose6_out1);
        let reshape68_out1 = linear92_out1.reshape([1, -1, 8, 128]);
        let constant176_out1 = self.constant176.val();
        let add77_out1 = reshape65_out1
            .clone()
            .add((constant176_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose142_out1 = add77_out1.permute([0, 2, 1, 3]);
        let constant177_out1 = self.constant177.val();
        let add78_out1 = reshape65_out1
            .add((constant177_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose143_out1 = add78_out1.permute([0, 2, 1, 3]);
        let transpose144_out1 = reshape68_out1.permute([0, 2, 3, 1]);
        let matmul120_out1 = transpose143_out1.matmul(transpose144_out1);
        let pad19_out1 = matmul120_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape69_out1 = pad19_out1.reshape(concat16_out1);
        let slice44_out1 = reshape69_out1.slice(s![.., .., 1.., ..]);
        let reshape70_out1 = slice44_out1.reshape(concat17_out1);
        let transpose145_out1 = reshape70_out1.permute([3, 0, 1, 2]);
        let slice45_out1 = transpose145_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose146_out1 = slice45_out1.permute([1, 2, 3, 0]);
        let mul85_out1 = transpose146_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where28_out1 = mul85_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention10_out1,) = {
            let q = transpose142_out1;
            let k = transpose140_out1;
            let v = transpose141_out1;
            let score_bias = where28_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention10_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention10_out1 = {
                let dims = attention10_out1.dims();
                attention10_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention10_out1,)
        };
        let where29_out1 = attention10_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose147_out1 = where29_out1.permute([0, 2, 1, 3]);
        let reshape71_out1 = transpose147_out1.reshape([1, -1, 1024]);
        let linear93_out1 = self.linear93.forward(reshape71_out1);
        let add79_out1 = add76_out1.add(linear93_out1);
        (add79_out1, gather20_out1, concat44_out1, slice42_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule13 {
    layernormalization57: LayerNorm,
    squeeze20: burn::module::Param<Tensor<2>>,
    conv1d10: Conv1d,
    layernormalization58: LayerNorm,
    squeeze21: burn::module::Param<Tensor<2>>,
    layernormalization59: LayerNorm,
    linear94: Linear,
    linear95: Linear,
    layernormalization60: LayerNorm,
    layernormalization61: LayerNorm,
    linear96: Linear,
    linear97: Linear,
    layernormalization62: LayerNorm,
    linear98: Linear,
    linear99: Linear,
    linear100: Linear,
    linear101: Linear,
    constant193: burn::module::Param<Tensor<2>>,
    constant194: burn::module::Param<Tensor<2>>,
    linear102: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule13 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization57 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze20: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d10 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization58 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze21: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization59 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear94 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear95 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization60 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization61 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear96 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear97 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization62 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear98 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear99 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear100 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear101 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant193: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant194: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear102 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization57,
            squeeze20,
            conv1d10,
            layernormalization58,
            squeeze21,
            layernormalization59,
            linear94,
            linear95,
            layernormalization60,
            layernormalization61,
            linear96,
            linear97,
            layernormalization62,
            linear98,
            linear99,
            linear100,
            linear101,
            constant193,
            constant194,
            linear102,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add79_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather20_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization57_out1 = {
            let dtype = add79_out1.clone().dtype();
            self.layernormalization57
                .forward(add79_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze20_out1 = self.squeeze20.val();
        let transpose148_out1 = squeeze20_out1.permute([1, 0]);
        let matmul122_out1 = layernormalization57_out1
            .matmul(transpose148_out1.unsqueeze::<3usize>());
        let split_tensors = matmul122_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split10_out1, split10_out2] = split_tensors.try_into().unwrap();
        let sigmoid38_out1 = burn::tensor::activation::sigmoid(split10_out2);
        let mul86_out1 = split10_out1.mul(sigmoid38_out1);
        let transpose149_out1 = mul86_out1.permute([0, 2, 1]);
        let where30_out1 = transpose149_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad20_out1 = where30_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat45_out1 = burn::tensor::Tensor::cat(
            [gather20_out1, pad20_out1].into(),
            2,
        );
        let slice46_out1 = concat45_out1.clone().slice(s![.., .., - 8..]);
        let conv1d10_out1 = self.conv1d10.forward(concat45_out1);
        let transpose150_out1 = conv1d10_out1.permute([0, 2, 1]);
        let layernormalization58_out1 = {
            let dtype = transpose150_out1.dtype();
            self.layernormalization58
                .forward(transpose150_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid39_out1 = burn::tensor::activation::sigmoid(
            layernormalization58_out1.clone(),
        );
        let mul87_out1 = layernormalization58_out1.mul(sigmoid39_out1);
        let squeeze21_out1 = self.squeeze21.val();
        let transpose151_out1 = squeeze21_out1.permute([1, 0]);
        let matmul123_out1 = mul87_out1.matmul(transpose151_out1.unsqueeze::<3usize>());
        let add80_out1 = add79_out1.add(matmul123_out1);
        let layernormalization59_out1 = {
            let dtype = add80_out1.clone().dtype();
            self.layernormalization59
                .forward(add80_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear94_out1 = self.linear94.forward(layernormalization59_out1);
        let sigmoid40_out1 = burn::tensor::activation::sigmoid(linear94_out1.clone());
        let mul88_out1 = linear94_out1.mul(sigmoid40_out1);
        let linear95_out1 = self.linear95.forward(mul88_out1);
        let mul89_out1 = linear95_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add81_out1 = add80_out1.add(mul89_out1);
        let layernormalization60_out1 = {
            let dtype = add81_out1.dtype();
            self.layernormalization60
                .forward(add81_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather21_out1 = {
            let sliced = transpose1_out1.slice(s![10, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather22_out1 = {
            let sliced = transpose2_out1.slice(s![10, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization61_out1 = {
            let dtype = layernormalization60_out1.clone().dtype();
            self.layernormalization61
                .forward(
                    layernormalization60_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear96_out1 = self.linear96.forward(layernormalization61_out1);
        let sigmoid41_out1 = burn::tensor::activation::sigmoid(linear96_out1.clone());
        let mul90_out1 = linear96_out1.mul(sigmoid41_out1);
        let linear97_out1 = self.linear97.forward(mul90_out1);
        let mul91_out1 = linear97_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add82_out1 = layernormalization60_out1.add(mul91_out1);
        let layernormalization62_out1 = {
            let dtype = add82_out1.clone().dtype();
            self.layernormalization62
                .forward(add82_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat46_out1 = burn::tensor::Tensor::cat(
            [gather21_out1.clone(), layernormalization62_out1.clone()].into(),
            1,
        );
        let transpose152_out1 = gather21_out1.permute([1, 0, 2]);
        let slice47_out1 = transpose152_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose153_out1 = slice47_out1.permute([1, 0, 2]);
        let concat47_out1 = burn::tensor::Tensor::cat(
            [transpose153_out1, layernormalization62_out1.clone()].into(),
            1,
        );
        let linear98_out1 = self.linear98.forward(layernormalization62_out1);
        let reshape72_out1 = linear98_out1.reshape([1, -1, 8, 128]);
        let linear99_out1 = self.linear99.forward(concat46_out1.clone());
        let reshape73_out1 = linear99_out1.reshape([1, -1, 8, 128]);
        let linear100_out1 = self.linear100.forward(concat46_out1);
        let reshape74_out1 = linear100_out1.reshape([1, -1, 8, 128]);
        let transpose154_out1 = reshape73_out1.permute([0, 2, 1, 3]);
        let transpose155_out1 = reshape74_out1.permute([0, 2, 1, 3]);
        let linear101_out1 = self.linear101.forward(transpose6_out1);
        let reshape75_out1 = linear101_out1.reshape([1, -1, 8, 128]);
        let constant193_out1 = self.constant193.val();
        let add83_out1 = reshape72_out1
            .clone()
            .add((constant193_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose156_out1 = add83_out1.permute([0, 2, 1, 3]);
        let constant194_out1 = self.constant194.val();
        let add84_out1 = reshape72_out1
            .add((constant194_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose157_out1 = add84_out1.permute([0, 2, 1, 3]);
        let transpose158_out1 = reshape75_out1.permute([0, 2, 3, 1]);
        let matmul132_out1 = transpose157_out1.matmul(transpose158_out1);
        let pad21_out1 = matmul132_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape76_out1 = pad21_out1.reshape(concat16_out1);
        let slice48_out1 = reshape76_out1.slice(s![.., .., 1.., ..]);
        let reshape77_out1 = slice48_out1.reshape(concat17_out1);
        let transpose159_out1 = reshape77_out1.permute([3, 0, 1, 2]);
        let slice49_out1 = transpose159_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose160_out1 = slice49_out1.permute([1, 2, 3, 0]);
        let mul92_out1 = transpose160_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where31_out1 = mul92_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention11_out1,) = {
            let q = transpose156_out1;
            let k = transpose154_out1;
            let v = transpose155_out1;
            let score_bias = where31_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention11_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention11_out1 = {
                let dims = attention11_out1.dims();
                attention11_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention11_out1,)
        };
        let where32_out1 = attention11_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose161_out1 = where32_out1.permute([0, 2, 1, 3]);
        let reshape78_out1 = transpose161_out1.reshape([1, -1, 1024]);
        let linear102_out1 = self.linear102.forward(reshape78_out1);
        let add85_out1 = add82_out1.add(linear102_out1);
        (add85_out1, gather22_out1, concat47_out1, slice46_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule14 {
    layernormalization63: LayerNorm,
    squeeze22: burn::module::Param<Tensor<2>>,
    conv1d11: Conv1d,
    layernormalization64: LayerNorm,
    squeeze23: burn::module::Param<Tensor<2>>,
    layernormalization65: LayerNorm,
    linear103: Linear,
    linear104: Linear,
    layernormalization66: LayerNorm,
    layernormalization67: LayerNorm,
    linear105: Linear,
    linear106: Linear,
    layernormalization68: LayerNorm,
    linear107: Linear,
    linear108: Linear,
    linear109: Linear,
    linear110: Linear,
    constant210: burn::module::Param<Tensor<2>>,
    constant211: burn::module::Param<Tensor<2>>,
    linear111: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule14 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization63 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze22: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d11 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization64 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze23: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization65 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear103 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear104 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization66 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization67 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear105 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear106 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization68 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear107 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear108 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear109 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear110 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant210: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant211: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear111 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization63,
            squeeze22,
            conv1d11,
            layernormalization64,
            squeeze23,
            layernormalization65,
            linear103,
            linear104,
            layernormalization66,
            layernormalization67,
            linear105,
            linear106,
            layernormalization68,
            linear107,
            linear108,
            linear109,
            linear110,
            constant210,
            constant211,
            linear111,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add85_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather22_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization63_out1 = {
            let dtype = add85_out1.clone().dtype();
            self.layernormalization63
                .forward(add85_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze22_out1 = self.squeeze22.val();
        let transpose162_out1 = squeeze22_out1.permute([1, 0]);
        let matmul134_out1 = layernormalization63_out1
            .matmul(transpose162_out1.unsqueeze::<3usize>());
        let split_tensors = matmul134_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split11_out1, split11_out2] = split_tensors.try_into().unwrap();
        let sigmoid42_out1 = burn::tensor::activation::sigmoid(split11_out2);
        let mul93_out1 = split11_out1.mul(sigmoid42_out1);
        let transpose163_out1 = mul93_out1.permute([0, 2, 1]);
        let where33_out1 = transpose163_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad22_out1 = where33_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat48_out1 = burn::tensor::Tensor::cat(
            [gather22_out1, pad22_out1].into(),
            2,
        );
        let slice50_out1 = concat48_out1.clone().slice(s![.., .., - 8..]);
        let conv1d11_out1 = self.conv1d11.forward(concat48_out1);
        let transpose164_out1 = conv1d11_out1.permute([0, 2, 1]);
        let layernormalization64_out1 = {
            let dtype = transpose164_out1.dtype();
            self.layernormalization64
                .forward(transpose164_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid43_out1 = burn::tensor::activation::sigmoid(
            layernormalization64_out1.clone(),
        );
        let mul94_out1 = layernormalization64_out1.mul(sigmoid43_out1);
        let squeeze23_out1 = self.squeeze23.val();
        let transpose165_out1 = squeeze23_out1.permute([1, 0]);
        let matmul135_out1 = mul94_out1.matmul(transpose165_out1.unsqueeze::<3usize>());
        let add86_out1 = add85_out1.add(matmul135_out1);
        let layernormalization65_out1 = {
            let dtype = add86_out1.clone().dtype();
            self.layernormalization65
                .forward(add86_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear103_out1 = self.linear103.forward(layernormalization65_out1);
        let sigmoid44_out1 = burn::tensor::activation::sigmoid(linear103_out1.clone());
        let mul95_out1 = linear103_out1.mul(sigmoid44_out1);
        let linear104_out1 = self.linear104.forward(mul95_out1);
        let mul96_out1 = linear104_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add87_out1 = add86_out1.add(mul96_out1);
        let layernormalization66_out1 = {
            let dtype = add87_out1.dtype();
            self.layernormalization66
                .forward(add87_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather23_out1 = {
            let sliced = transpose1_out1.slice(s![11, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather24_out1 = {
            let sliced = transpose2_out1.slice(s![11, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization67_out1 = {
            let dtype = layernormalization66_out1.clone().dtype();
            self.layernormalization67
                .forward(
                    layernormalization66_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear105_out1 = self.linear105.forward(layernormalization67_out1);
        let sigmoid45_out1 = burn::tensor::activation::sigmoid(linear105_out1.clone());
        let mul97_out1 = linear105_out1.mul(sigmoid45_out1);
        let linear106_out1 = self.linear106.forward(mul97_out1);
        let mul98_out1 = linear106_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add88_out1 = layernormalization66_out1.add(mul98_out1);
        let layernormalization68_out1 = {
            let dtype = add88_out1.clone().dtype();
            self.layernormalization68
                .forward(add88_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat49_out1 = burn::tensor::Tensor::cat(
            [gather23_out1.clone(), layernormalization68_out1.clone()].into(),
            1,
        );
        let transpose166_out1 = gather23_out1.permute([1, 0, 2]);
        let slice51_out1 = transpose166_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose167_out1 = slice51_out1.permute([1, 0, 2]);
        let concat50_out1 = burn::tensor::Tensor::cat(
            [transpose167_out1, layernormalization68_out1.clone()].into(),
            1,
        );
        let linear107_out1 = self.linear107.forward(layernormalization68_out1);
        let reshape79_out1 = linear107_out1.reshape([1, -1, 8, 128]);
        let linear108_out1 = self.linear108.forward(concat49_out1.clone());
        let reshape80_out1 = linear108_out1.reshape([1, -1, 8, 128]);
        let linear109_out1 = self.linear109.forward(concat49_out1);
        let reshape81_out1 = linear109_out1.reshape([1, -1, 8, 128]);
        let transpose168_out1 = reshape80_out1.permute([0, 2, 1, 3]);
        let transpose169_out1 = reshape81_out1.permute([0, 2, 1, 3]);
        let linear110_out1 = self.linear110.forward(transpose6_out1);
        let reshape82_out1 = linear110_out1.reshape([1, -1, 8, 128]);
        let constant210_out1 = self.constant210.val();
        let add89_out1 = reshape79_out1
            .clone()
            .add((constant210_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose170_out1 = add89_out1.permute([0, 2, 1, 3]);
        let constant211_out1 = self.constant211.val();
        let add90_out1 = reshape79_out1
            .add((constant211_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose171_out1 = add90_out1.permute([0, 2, 1, 3]);
        let transpose172_out1 = reshape82_out1.permute([0, 2, 3, 1]);
        let matmul144_out1 = transpose171_out1.matmul(transpose172_out1);
        let pad23_out1 = matmul144_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape83_out1 = pad23_out1.reshape(concat16_out1);
        let slice52_out1 = reshape83_out1.slice(s![.., .., 1.., ..]);
        let reshape84_out1 = slice52_out1.reshape(concat17_out1);
        let transpose173_out1 = reshape84_out1.permute([3, 0, 1, 2]);
        let slice53_out1 = transpose173_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose174_out1 = slice53_out1.permute([1, 2, 3, 0]);
        let mul99_out1 = transpose174_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where34_out1 = mul99_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention12_out1,) = {
            let q = transpose170_out1;
            let k = transpose168_out1;
            let v = transpose169_out1;
            let score_bias = where34_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention12_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention12_out1 = {
                let dims = attention12_out1.dims();
                attention12_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention12_out1,)
        };
        let where35_out1 = attention12_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose175_out1 = where35_out1.permute([0, 2, 1, 3]);
        let reshape85_out1 = transpose175_out1.reshape([1, -1, 1024]);
        let linear111_out1 = self.linear111.forward(reshape85_out1);
        let add91_out1 = add88_out1.add(linear111_out1);
        (add91_out1, gather24_out1, concat50_out1, slice50_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule15 {
    layernormalization69: LayerNorm,
    squeeze24: burn::module::Param<Tensor<2>>,
    conv1d12: Conv1d,
    layernormalization70: LayerNorm,
    squeeze25: burn::module::Param<Tensor<2>>,
    layernormalization71: LayerNorm,
    linear112: Linear,
    linear113: Linear,
    layernormalization72: LayerNorm,
    layernormalization73: LayerNorm,
    linear114: Linear,
    linear115: Linear,
    layernormalization74: LayerNorm,
    linear116: Linear,
    linear117: Linear,
    linear118: Linear,
    linear119: Linear,
    constant227: burn::module::Param<Tensor<2>>,
    constant228: burn::module::Param<Tensor<2>>,
    linear120: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule15 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization69 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze24: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d12 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization70 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze25: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization71 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear112 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear113 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization72 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization73 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear114 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear115 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization74 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear116 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear117 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear118 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear119 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant227: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant228: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear120 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization69,
            squeeze24,
            conv1d12,
            layernormalization70,
            squeeze25,
            layernormalization71,
            linear112,
            linear113,
            layernormalization72,
            layernormalization73,
            linear114,
            linear115,
            layernormalization74,
            linear116,
            linear117,
            linear118,
            linear119,
            constant227,
            constant228,
            linear120,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add91_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather24_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization69_out1 = {
            let dtype = add91_out1.clone().dtype();
            self.layernormalization69
                .forward(add91_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze24_out1 = self.squeeze24.val();
        let transpose176_out1 = squeeze24_out1.permute([1, 0]);
        let matmul146_out1 = layernormalization69_out1
            .matmul(transpose176_out1.unsqueeze::<3usize>());
        let split_tensors = matmul146_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split12_out1, split12_out2] = split_tensors.try_into().unwrap();
        let sigmoid46_out1 = burn::tensor::activation::sigmoid(split12_out2);
        let mul100_out1 = split12_out1.mul(sigmoid46_out1);
        let transpose177_out1 = mul100_out1.permute([0, 2, 1]);
        let where36_out1 = transpose177_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad24_out1 = where36_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat51_out1 = burn::tensor::Tensor::cat(
            [gather24_out1, pad24_out1].into(),
            2,
        );
        let slice54_out1 = concat51_out1.clone().slice(s![.., .., - 8..]);
        let conv1d12_out1 = self.conv1d12.forward(concat51_out1);
        let transpose178_out1 = conv1d12_out1.permute([0, 2, 1]);
        let layernormalization70_out1 = {
            let dtype = transpose178_out1.dtype();
            self.layernormalization70
                .forward(transpose178_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid47_out1 = burn::tensor::activation::sigmoid(
            layernormalization70_out1.clone(),
        );
        let mul101_out1 = layernormalization70_out1.mul(sigmoid47_out1);
        let squeeze25_out1 = self.squeeze25.val();
        let transpose179_out1 = squeeze25_out1.permute([1, 0]);
        let matmul147_out1 = mul101_out1.matmul(transpose179_out1.unsqueeze::<3usize>());
        let add92_out1 = add91_out1.add(matmul147_out1);
        let layernormalization71_out1 = {
            let dtype = add92_out1.clone().dtype();
            self.layernormalization71
                .forward(add92_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear112_out1 = self.linear112.forward(layernormalization71_out1);
        let sigmoid48_out1 = burn::tensor::activation::sigmoid(linear112_out1.clone());
        let mul102_out1 = linear112_out1.mul(sigmoid48_out1);
        let linear113_out1 = self.linear113.forward(mul102_out1);
        let mul103_out1 = linear113_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add93_out1 = add92_out1.add(mul103_out1);
        let layernormalization72_out1 = {
            let dtype = add93_out1.dtype();
            self.layernormalization72
                .forward(add93_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather25_out1 = {
            let sliced = transpose1_out1.slice(s![12, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather26_out1 = {
            let sliced = transpose2_out1.slice(s![12, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization73_out1 = {
            let dtype = layernormalization72_out1.clone().dtype();
            self.layernormalization73
                .forward(
                    layernormalization72_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear114_out1 = self.linear114.forward(layernormalization73_out1);
        let sigmoid49_out1 = burn::tensor::activation::sigmoid(linear114_out1.clone());
        let mul104_out1 = linear114_out1.mul(sigmoid49_out1);
        let linear115_out1 = self.linear115.forward(mul104_out1);
        let mul105_out1 = linear115_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add94_out1 = layernormalization72_out1.add(mul105_out1);
        let layernormalization74_out1 = {
            let dtype = add94_out1.clone().dtype();
            self.layernormalization74
                .forward(add94_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat52_out1 = burn::tensor::Tensor::cat(
            [gather25_out1.clone(), layernormalization74_out1.clone()].into(),
            1,
        );
        let transpose180_out1 = gather25_out1.permute([1, 0, 2]);
        let slice55_out1 = transpose180_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose181_out1 = slice55_out1.permute([1, 0, 2]);
        let concat53_out1 = burn::tensor::Tensor::cat(
            [transpose181_out1, layernormalization74_out1.clone()].into(),
            1,
        );
        let linear116_out1 = self.linear116.forward(layernormalization74_out1);
        let reshape86_out1 = linear116_out1.reshape([1, -1, 8, 128]);
        let linear117_out1 = self.linear117.forward(concat52_out1.clone());
        let reshape87_out1 = linear117_out1.reshape([1, -1, 8, 128]);
        let linear118_out1 = self.linear118.forward(concat52_out1);
        let reshape88_out1 = linear118_out1.reshape([1, -1, 8, 128]);
        let transpose182_out1 = reshape87_out1.permute([0, 2, 1, 3]);
        let transpose183_out1 = reshape88_out1.permute([0, 2, 1, 3]);
        let linear119_out1 = self.linear119.forward(transpose6_out1);
        let reshape89_out1 = linear119_out1.reshape([1, -1, 8, 128]);
        let constant227_out1 = self.constant227.val();
        let add95_out1 = reshape86_out1
            .clone()
            .add((constant227_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose184_out1 = add95_out1.permute([0, 2, 1, 3]);
        let constant228_out1 = self.constant228.val();
        let add96_out1 = reshape86_out1
            .add((constant228_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose185_out1 = add96_out1.permute([0, 2, 1, 3]);
        let transpose186_out1 = reshape89_out1.permute([0, 2, 3, 1]);
        let matmul156_out1 = transpose185_out1.matmul(transpose186_out1);
        let pad25_out1 = matmul156_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape90_out1 = pad25_out1.reshape(concat16_out1);
        let slice56_out1 = reshape90_out1.slice(s![.., .., 1.., ..]);
        let reshape91_out1 = slice56_out1.reshape(concat17_out1);
        let transpose187_out1 = reshape91_out1.permute([3, 0, 1, 2]);
        let slice57_out1 = transpose187_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose188_out1 = slice57_out1.permute([1, 2, 3, 0]);
        let mul106_out1 = transpose188_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where37_out1 = mul106_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention13_out1,) = {
            let q = transpose184_out1;
            let k = transpose182_out1;
            let v = transpose183_out1;
            let score_bias = where37_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention13_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention13_out1 = {
                let dims = attention13_out1.dims();
                attention13_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention13_out1,)
        };
        let where38_out1 = attention13_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose189_out1 = where38_out1.permute([0, 2, 1, 3]);
        let reshape92_out1 = transpose189_out1.reshape([1, -1, 1024]);
        let linear120_out1 = self.linear120.forward(reshape92_out1);
        let add97_out1 = add94_out1.add(linear120_out1);
        (add97_out1, gather26_out1, concat53_out1, slice54_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule16 {
    layernormalization75: LayerNorm,
    squeeze26: burn::module::Param<Tensor<2>>,
    conv1d13: Conv1d,
    layernormalization76: LayerNorm,
    squeeze27: burn::module::Param<Tensor<2>>,
    layernormalization77: LayerNorm,
    linear121: Linear,
    linear122: Linear,
    layernormalization78: LayerNorm,
    layernormalization79: LayerNorm,
    linear123: Linear,
    linear124: Linear,
    layernormalization80: LayerNorm,
    linear125: Linear,
    linear126: Linear,
    linear127: Linear,
    linear128: Linear,
    constant244: burn::module::Param<Tensor<2>>,
    constant245: burn::module::Param<Tensor<2>>,
    linear129: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule16 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization75 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze26: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d13 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization76 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze27: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization77 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear121 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear122 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization78 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization79 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear123 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear124 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization80 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear125 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear126 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear127 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear128 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant244: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant245: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear129 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization75,
            squeeze26,
            conv1d13,
            layernormalization76,
            squeeze27,
            layernormalization77,
            linear121,
            linear122,
            layernormalization78,
            layernormalization79,
            linear123,
            linear124,
            layernormalization80,
            linear125,
            linear126,
            linear127,
            linear128,
            constant244,
            constant245,
            linear129,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add97_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather26_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization75_out1 = {
            let dtype = add97_out1.clone().dtype();
            self.layernormalization75
                .forward(add97_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze26_out1 = self.squeeze26.val();
        let transpose190_out1 = squeeze26_out1.permute([1, 0]);
        let matmul158_out1 = layernormalization75_out1
            .matmul(transpose190_out1.unsqueeze::<3usize>());
        let split_tensors = matmul158_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split13_out1, split13_out2] = split_tensors.try_into().unwrap();
        let sigmoid50_out1 = burn::tensor::activation::sigmoid(split13_out2);
        let mul107_out1 = split13_out1.mul(sigmoid50_out1);
        let transpose191_out1 = mul107_out1.permute([0, 2, 1]);
        let where39_out1 = transpose191_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad26_out1 = where39_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat54_out1 = burn::tensor::Tensor::cat(
            [gather26_out1, pad26_out1].into(),
            2,
        );
        let slice58_out1 = concat54_out1.clone().slice(s![.., .., - 8..]);
        let conv1d13_out1 = self.conv1d13.forward(concat54_out1);
        let transpose192_out1 = conv1d13_out1.permute([0, 2, 1]);
        let layernormalization76_out1 = {
            let dtype = transpose192_out1.dtype();
            self.layernormalization76
                .forward(transpose192_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid51_out1 = burn::tensor::activation::sigmoid(
            layernormalization76_out1.clone(),
        );
        let mul108_out1 = layernormalization76_out1.mul(sigmoid51_out1);
        let squeeze27_out1 = self.squeeze27.val();
        let transpose193_out1 = squeeze27_out1.permute([1, 0]);
        let matmul159_out1 = mul108_out1.matmul(transpose193_out1.unsqueeze::<3usize>());
        let add98_out1 = add97_out1.add(matmul159_out1);
        let layernormalization77_out1 = {
            let dtype = add98_out1.clone().dtype();
            self.layernormalization77
                .forward(add98_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear121_out1 = self.linear121.forward(layernormalization77_out1);
        let sigmoid52_out1 = burn::tensor::activation::sigmoid(linear121_out1.clone());
        let mul109_out1 = linear121_out1.mul(sigmoid52_out1);
        let linear122_out1 = self.linear122.forward(mul109_out1);
        let mul110_out1 = linear122_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add99_out1 = add98_out1.add(mul110_out1);
        let layernormalization78_out1 = {
            let dtype = add99_out1.dtype();
            self.layernormalization78
                .forward(add99_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather27_out1 = {
            let sliced = transpose1_out1.slice(s![13, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather28_out1 = {
            let sliced = transpose2_out1.slice(s![13, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization79_out1 = {
            let dtype = layernormalization78_out1.clone().dtype();
            self.layernormalization79
                .forward(
                    layernormalization78_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear123_out1 = self.linear123.forward(layernormalization79_out1);
        let sigmoid53_out1 = burn::tensor::activation::sigmoid(linear123_out1.clone());
        let mul111_out1 = linear123_out1.mul(sigmoid53_out1);
        let linear124_out1 = self.linear124.forward(mul111_out1);
        let mul112_out1 = linear124_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add100_out1 = layernormalization78_out1.add(mul112_out1);
        let layernormalization80_out1 = {
            let dtype = add100_out1.clone().dtype();
            self.layernormalization80
                .forward(add100_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat55_out1 = burn::tensor::Tensor::cat(
            [gather27_out1.clone(), layernormalization80_out1.clone()].into(),
            1,
        );
        let transpose194_out1 = gather27_out1.permute([1, 0, 2]);
        let slice59_out1 = transpose194_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose195_out1 = slice59_out1.permute([1, 0, 2]);
        let concat56_out1 = burn::tensor::Tensor::cat(
            [transpose195_out1, layernormalization80_out1.clone()].into(),
            1,
        );
        let linear125_out1 = self.linear125.forward(layernormalization80_out1);
        let reshape93_out1 = linear125_out1.reshape([1, -1, 8, 128]);
        let linear126_out1 = self.linear126.forward(concat55_out1.clone());
        let reshape94_out1 = linear126_out1.reshape([1, -1, 8, 128]);
        let linear127_out1 = self.linear127.forward(concat55_out1);
        let reshape95_out1 = linear127_out1.reshape([1, -1, 8, 128]);
        let transpose196_out1 = reshape94_out1.permute([0, 2, 1, 3]);
        let transpose197_out1 = reshape95_out1.permute([0, 2, 1, 3]);
        let linear128_out1 = self.linear128.forward(transpose6_out1);
        let reshape96_out1 = linear128_out1.reshape([1, -1, 8, 128]);
        let constant244_out1 = self.constant244.val();
        let add101_out1 = reshape93_out1
            .clone()
            .add((constant244_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose198_out1 = add101_out1.permute([0, 2, 1, 3]);
        let constant245_out1 = self.constant245.val();
        let add102_out1 = reshape93_out1
            .add((constant245_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose199_out1 = add102_out1.permute([0, 2, 1, 3]);
        let transpose200_out1 = reshape96_out1.permute([0, 2, 3, 1]);
        let matmul168_out1 = transpose199_out1.matmul(transpose200_out1);
        let pad27_out1 = matmul168_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape97_out1 = pad27_out1.reshape(concat16_out1);
        let slice60_out1 = reshape97_out1.slice(s![.., .., 1.., ..]);
        let reshape98_out1 = slice60_out1.reshape(concat17_out1);
        let transpose201_out1 = reshape98_out1.permute([3, 0, 1, 2]);
        let slice61_out1 = transpose201_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose202_out1 = slice61_out1.permute([1, 2, 3, 0]);
        let mul113_out1 = transpose202_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where40_out1 = mul113_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention14_out1,) = {
            let q = transpose198_out1;
            let k = transpose196_out1;
            let v = transpose197_out1;
            let score_bias = where40_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention14_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention14_out1 = {
                let dims = attention14_out1.dims();
                attention14_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention14_out1,)
        };
        let where41_out1 = attention14_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose203_out1 = where41_out1.permute([0, 2, 1, 3]);
        let reshape99_out1 = transpose203_out1.reshape([1, -1, 1024]);
        let linear129_out1 = self.linear129.forward(reshape99_out1);
        let add103_out1 = add100_out1.add(linear129_out1);
        (add103_out1, gather28_out1, concat56_out1, slice58_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule17 {
    layernormalization81: LayerNorm,
    squeeze28: burn::module::Param<Tensor<2>>,
    conv1d14: Conv1d,
    layernormalization82: LayerNorm,
    squeeze29: burn::module::Param<Tensor<2>>,
    layernormalization83: LayerNorm,
    linear130: Linear,
    linear131: Linear,
    layernormalization84: LayerNorm,
    layernormalization85: LayerNorm,
    linear132: Linear,
    linear133: Linear,
    layernormalization86: LayerNorm,
    linear134: Linear,
    linear135: Linear,
    linear136: Linear,
    linear137: Linear,
    constant261: burn::module::Param<Tensor<2>>,
    constant262: burn::module::Param<Tensor<2>>,
    linear138: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule17 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization81 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze28: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d14 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization82 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze29: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization83 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear130 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear131 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization84 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization85 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear132 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear133 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization86 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear134 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear135 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear136 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear137 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant261: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant262: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear138 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization81,
            squeeze28,
            conv1d14,
            layernormalization82,
            squeeze29,
            layernormalization83,
            linear130,
            linear131,
            layernormalization84,
            layernormalization85,
            linear132,
            linear133,
            layernormalization86,
            linear134,
            linear135,
            linear136,
            linear137,
            constant261,
            constant262,
            linear138,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add103_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather28_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization81_out1 = {
            let dtype = add103_out1.clone().dtype();
            self.layernormalization81
                .forward(add103_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze28_out1 = self.squeeze28.val();
        let transpose204_out1 = squeeze28_out1.permute([1, 0]);
        let matmul170_out1 = layernormalization81_out1
            .matmul(transpose204_out1.unsqueeze::<3usize>());
        let split_tensors = matmul170_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split14_out1, split14_out2] = split_tensors.try_into().unwrap();
        let sigmoid54_out1 = burn::tensor::activation::sigmoid(split14_out2);
        let mul114_out1 = split14_out1.mul(sigmoid54_out1);
        let transpose205_out1 = mul114_out1.permute([0, 2, 1]);
        let where42_out1 = transpose205_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad28_out1 = where42_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat57_out1 = burn::tensor::Tensor::cat(
            [gather28_out1, pad28_out1].into(),
            2,
        );
        let slice62_out1 = concat57_out1.clone().slice(s![.., .., - 8..]);
        let conv1d14_out1 = self.conv1d14.forward(concat57_out1);
        let transpose206_out1 = conv1d14_out1.permute([0, 2, 1]);
        let layernormalization82_out1 = {
            let dtype = transpose206_out1.dtype();
            self.layernormalization82
                .forward(transpose206_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid55_out1 = burn::tensor::activation::sigmoid(
            layernormalization82_out1.clone(),
        );
        let mul115_out1 = layernormalization82_out1.mul(sigmoid55_out1);
        let squeeze29_out1 = self.squeeze29.val();
        let transpose207_out1 = squeeze29_out1.permute([1, 0]);
        let matmul171_out1 = mul115_out1.matmul(transpose207_out1.unsqueeze::<3usize>());
        let add104_out1 = add103_out1.add(matmul171_out1);
        let layernormalization83_out1 = {
            let dtype = add104_out1.clone().dtype();
            self.layernormalization83
                .forward(add104_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear130_out1 = self.linear130.forward(layernormalization83_out1);
        let sigmoid56_out1 = burn::tensor::activation::sigmoid(linear130_out1.clone());
        let mul116_out1 = linear130_out1.mul(sigmoid56_out1);
        let linear131_out1 = self.linear131.forward(mul116_out1);
        let mul117_out1 = linear131_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add105_out1 = add104_out1.add(mul117_out1);
        let layernormalization84_out1 = {
            let dtype = add105_out1.dtype();
            self.layernormalization84
                .forward(add105_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather29_out1 = {
            let sliced = transpose1_out1.slice(s![14, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather30_out1 = {
            let sliced = transpose2_out1.slice(s![14, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization85_out1 = {
            let dtype = layernormalization84_out1.clone().dtype();
            self.layernormalization85
                .forward(
                    layernormalization84_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear132_out1 = self.linear132.forward(layernormalization85_out1);
        let sigmoid57_out1 = burn::tensor::activation::sigmoid(linear132_out1.clone());
        let mul118_out1 = linear132_out1.mul(sigmoid57_out1);
        let linear133_out1 = self.linear133.forward(mul118_out1);
        let mul119_out1 = linear133_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add106_out1 = layernormalization84_out1.add(mul119_out1);
        let layernormalization86_out1 = {
            let dtype = add106_out1.clone().dtype();
            self.layernormalization86
                .forward(add106_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat58_out1 = burn::tensor::Tensor::cat(
            [gather29_out1.clone(), layernormalization86_out1.clone()].into(),
            1,
        );
        let transpose208_out1 = gather29_out1.permute([1, 0, 2]);
        let slice63_out1 = transpose208_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose209_out1 = slice63_out1.permute([1, 0, 2]);
        let concat59_out1 = burn::tensor::Tensor::cat(
            [transpose209_out1, layernormalization86_out1.clone()].into(),
            1,
        );
        let linear134_out1 = self.linear134.forward(layernormalization86_out1);
        let reshape100_out1 = linear134_out1.reshape([1, -1, 8, 128]);
        let linear135_out1 = self.linear135.forward(concat58_out1.clone());
        let reshape101_out1 = linear135_out1.reshape([1, -1, 8, 128]);
        let linear136_out1 = self.linear136.forward(concat58_out1);
        let reshape102_out1 = linear136_out1.reshape([1, -1, 8, 128]);
        let transpose210_out1 = reshape101_out1.permute([0, 2, 1, 3]);
        let transpose211_out1 = reshape102_out1.permute([0, 2, 1, 3]);
        let linear137_out1 = self.linear137.forward(transpose6_out1);
        let reshape103_out1 = linear137_out1.reshape([1, -1, 8, 128]);
        let constant261_out1 = self.constant261.val();
        let add107_out1 = reshape100_out1
            .clone()
            .add((constant261_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose212_out1 = add107_out1.permute([0, 2, 1, 3]);
        let constant262_out1 = self.constant262.val();
        let add108_out1 = reshape100_out1
            .add((constant262_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose213_out1 = add108_out1.permute([0, 2, 1, 3]);
        let transpose214_out1 = reshape103_out1.permute([0, 2, 3, 1]);
        let matmul180_out1 = transpose213_out1.matmul(transpose214_out1);
        let pad29_out1 = matmul180_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape104_out1 = pad29_out1.reshape(concat16_out1);
        let slice64_out1 = reshape104_out1.slice(s![.., .., 1.., ..]);
        let reshape105_out1 = slice64_out1.reshape(concat17_out1);
        let transpose215_out1 = reshape105_out1.permute([3, 0, 1, 2]);
        let slice65_out1 = transpose215_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose216_out1 = slice65_out1.permute([1, 2, 3, 0]);
        let mul120_out1 = transpose216_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where43_out1 = mul120_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention15_out1,) = {
            let q = transpose212_out1;
            let k = transpose210_out1;
            let v = transpose211_out1;
            let score_bias = where43_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention15_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention15_out1 = {
                let dims = attention15_out1.dims();
                attention15_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention15_out1,)
        };
        let where44_out1 = attention15_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose217_out1 = where44_out1.permute([0, 2, 1, 3]);
        let reshape106_out1 = transpose217_out1.reshape([1, -1, 1024]);
        let linear138_out1 = self.linear138.forward(reshape106_out1);
        let add109_out1 = add106_out1.add(linear138_out1);
        (add109_out1, gather30_out1, concat59_out1, slice62_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule18 {
    layernormalization87: LayerNorm,
    squeeze30: burn::module::Param<Tensor<2>>,
    conv1d15: Conv1d,
    layernormalization88: LayerNorm,
    squeeze31: burn::module::Param<Tensor<2>>,
    layernormalization89: LayerNorm,
    linear139: Linear,
    linear140: Linear,
    layernormalization90: LayerNorm,
    layernormalization91: LayerNorm,
    linear141: Linear,
    linear142: Linear,
    layernormalization92: LayerNorm,
    linear143: Linear,
    linear144: Linear,
    linear145: Linear,
    linear146: Linear,
    constant278: burn::module::Param<Tensor<2>>,
    constant279: burn::module::Param<Tensor<2>>,
    linear147: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule18 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization87 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze30: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d15 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization88 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze31: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization89 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear139 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear140 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization90 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization91 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear141 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear142 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization92 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear143 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear144 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear145 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear146 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant278: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant279: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear147 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization87,
            squeeze30,
            conv1d15,
            layernormalization88,
            squeeze31,
            layernormalization89,
            linear139,
            linear140,
            layernormalization90,
            layernormalization91,
            linear141,
            linear142,
            layernormalization92,
            linear143,
            linear144,
            linear145,
            linear146,
            constant278,
            constant279,
            linear147,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add109_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather30_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization87_out1 = {
            let dtype = add109_out1.clone().dtype();
            self.layernormalization87
                .forward(add109_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze30_out1 = self.squeeze30.val();
        let transpose218_out1 = squeeze30_out1.permute([1, 0]);
        let matmul182_out1 = layernormalization87_out1
            .matmul(transpose218_out1.unsqueeze::<3usize>());
        let split_tensors = matmul182_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split15_out1, split15_out2] = split_tensors.try_into().unwrap();
        let sigmoid58_out1 = burn::tensor::activation::sigmoid(split15_out2);
        let mul121_out1 = split15_out1.mul(sigmoid58_out1);
        let transpose219_out1 = mul121_out1.permute([0, 2, 1]);
        let where45_out1 = transpose219_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad30_out1 = where45_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat60_out1 = burn::tensor::Tensor::cat(
            [gather30_out1, pad30_out1].into(),
            2,
        );
        let slice66_out1 = concat60_out1.clone().slice(s![.., .., - 8..]);
        let conv1d15_out1 = self.conv1d15.forward(concat60_out1);
        let transpose220_out1 = conv1d15_out1.permute([0, 2, 1]);
        let layernormalization88_out1 = {
            let dtype = transpose220_out1.dtype();
            self.layernormalization88
                .forward(transpose220_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid59_out1 = burn::tensor::activation::sigmoid(
            layernormalization88_out1.clone(),
        );
        let mul122_out1 = layernormalization88_out1.mul(sigmoid59_out1);
        let squeeze31_out1 = self.squeeze31.val();
        let transpose221_out1 = squeeze31_out1.permute([1, 0]);
        let matmul183_out1 = mul122_out1.matmul(transpose221_out1.unsqueeze::<3usize>());
        let add110_out1 = add109_out1.add(matmul183_out1);
        let layernormalization89_out1 = {
            let dtype = add110_out1.clone().dtype();
            self.layernormalization89
                .forward(add110_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear139_out1 = self.linear139.forward(layernormalization89_out1);
        let sigmoid60_out1 = burn::tensor::activation::sigmoid(linear139_out1.clone());
        let mul123_out1 = linear139_out1.mul(sigmoid60_out1);
        let linear140_out1 = self.linear140.forward(mul123_out1);
        let mul124_out1 = linear140_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add111_out1 = add110_out1.add(mul124_out1);
        let layernormalization90_out1 = {
            let dtype = add111_out1.dtype();
            self.layernormalization90
                .forward(add111_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather31_out1 = {
            let sliced = transpose1_out1.slice(s![15, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather32_out1 = {
            let sliced = transpose2_out1.slice(s![15, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization91_out1 = {
            let dtype = layernormalization90_out1.clone().dtype();
            self.layernormalization91
                .forward(
                    layernormalization90_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear141_out1 = self.linear141.forward(layernormalization91_out1);
        let sigmoid61_out1 = burn::tensor::activation::sigmoid(linear141_out1.clone());
        let mul125_out1 = linear141_out1.mul(sigmoid61_out1);
        let linear142_out1 = self.linear142.forward(mul125_out1);
        let mul126_out1 = linear142_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add112_out1 = layernormalization90_out1.add(mul126_out1);
        let layernormalization92_out1 = {
            let dtype = add112_out1.clone().dtype();
            self.layernormalization92
                .forward(add112_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat61_out1 = burn::tensor::Tensor::cat(
            [gather31_out1.clone(), layernormalization92_out1.clone()].into(),
            1,
        );
        let transpose222_out1 = gather31_out1.permute([1, 0, 2]);
        let slice67_out1 = transpose222_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose223_out1 = slice67_out1.permute([1, 0, 2]);
        let concat62_out1 = burn::tensor::Tensor::cat(
            [transpose223_out1, layernormalization92_out1.clone()].into(),
            1,
        );
        let linear143_out1 = self.linear143.forward(layernormalization92_out1);
        let reshape107_out1 = linear143_out1.reshape([1, -1, 8, 128]);
        let linear144_out1 = self.linear144.forward(concat61_out1.clone());
        let reshape108_out1 = linear144_out1.reshape([1, -1, 8, 128]);
        let linear145_out1 = self.linear145.forward(concat61_out1);
        let reshape109_out1 = linear145_out1.reshape([1, -1, 8, 128]);
        let transpose224_out1 = reshape108_out1.permute([0, 2, 1, 3]);
        let transpose225_out1 = reshape109_out1.permute([0, 2, 1, 3]);
        let linear146_out1 = self.linear146.forward(transpose6_out1);
        let reshape110_out1 = linear146_out1.reshape([1, -1, 8, 128]);
        let constant278_out1 = self.constant278.val();
        let add113_out1 = reshape107_out1
            .clone()
            .add((constant278_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose226_out1 = add113_out1.permute([0, 2, 1, 3]);
        let constant279_out1 = self.constant279.val();
        let add114_out1 = reshape107_out1
            .add((constant279_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose227_out1 = add114_out1.permute([0, 2, 1, 3]);
        let transpose228_out1 = reshape110_out1.permute([0, 2, 3, 1]);
        let matmul192_out1 = transpose227_out1.matmul(transpose228_out1);
        let pad31_out1 = matmul192_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape111_out1 = pad31_out1.reshape(concat16_out1);
        let slice68_out1 = reshape111_out1.slice(s![.., .., 1.., ..]);
        let reshape112_out1 = slice68_out1.reshape(concat17_out1);
        let transpose229_out1 = reshape112_out1.permute([3, 0, 1, 2]);
        let slice69_out1 = transpose229_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose230_out1 = slice69_out1.permute([1, 2, 3, 0]);
        let mul127_out1 = transpose230_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where46_out1 = mul127_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention16_out1,) = {
            let q = transpose226_out1;
            let k = transpose224_out1;
            let v = transpose225_out1;
            let score_bias = where46_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention16_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention16_out1 = {
                let dims = attention16_out1.dims();
                attention16_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention16_out1,)
        };
        let where47_out1 = attention16_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose231_out1 = where47_out1.permute([0, 2, 1, 3]);
        let reshape113_out1 = transpose231_out1.reshape([1, -1, 1024]);
        let linear147_out1 = self.linear147.forward(reshape113_out1);
        let add115_out1 = add112_out1.add(linear147_out1);
        (add115_out1, gather32_out1, concat62_out1, slice66_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule19 {
    layernormalization93: LayerNorm,
    squeeze32: burn::module::Param<Tensor<2>>,
    conv1d16: Conv1d,
    layernormalization94: LayerNorm,
    squeeze33: burn::module::Param<Tensor<2>>,
    layernormalization95: LayerNorm,
    linear148: Linear,
    linear149: Linear,
    layernormalization96: LayerNorm,
    layernormalization97: LayerNorm,
    linear150: Linear,
    linear151: Linear,
    layernormalization98: LayerNorm,
    linear152: Linear,
    linear153: Linear,
    linear154: Linear,
    linear155: Linear,
    constant295: burn::module::Param<Tensor<2>>,
    constant296: burn::module::Param<Tensor<2>>,
    linear156: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule19 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization93 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze32: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d16 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization94 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze33: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization95 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear148 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear149 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization96 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization97 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear150 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear151 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization98 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear152 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear153 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear154 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear155 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant295: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant296: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear156 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization93,
            squeeze32,
            conv1d16,
            layernormalization94,
            squeeze33,
            layernormalization95,
            linear148,
            linear149,
            layernormalization96,
            layernormalization97,
            linear150,
            linear151,
            layernormalization98,
            linear152,
            linear153,
            linear154,
            linear155,
            constant295,
            constant296,
            linear156,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add115_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather32_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization93_out1 = {
            let dtype = add115_out1.clone().dtype();
            self.layernormalization93
                .forward(add115_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze32_out1 = self.squeeze32.val();
        let transpose232_out1 = squeeze32_out1.permute([1, 0]);
        let matmul194_out1 = layernormalization93_out1
            .matmul(transpose232_out1.unsqueeze::<3usize>());
        let split_tensors = matmul194_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split16_out1, split16_out2] = split_tensors.try_into().unwrap();
        let sigmoid62_out1 = burn::tensor::activation::sigmoid(split16_out2);
        let mul128_out1 = split16_out1.mul(sigmoid62_out1);
        let transpose233_out1 = mul128_out1.permute([0, 2, 1]);
        let where48_out1 = transpose233_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad32_out1 = where48_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat63_out1 = burn::tensor::Tensor::cat(
            [gather32_out1, pad32_out1].into(),
            2,
        );
        let slice70_out1 = concat63_out1.clone().slice(s![.., .., - 8..]);
        let conv1d16_out1 = self.conv1d16.forward(concat63_out1);
        let transpose234_out1 = conv1d16_out1.permute([0, 2, 1]);
        let layernormalization94_out1 = {
            let dtype = transpose234_out1.dtype();
            self.layernormalization94
                .forward(transpose234_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid63_out1 = burn::tensor::activation::sigmoid(
            layernormalization94_out1.clone(),
        );
        let mul129_out1 = layernormalization94_out1.mul(sigmoid63_out1);
        let squeeze33_out1 = self.squeeze33.val();
        let transpose235_out1 = squeeze33_out1.permute([1, 0]);
        let matmul195_out1 = mul129_out1.matmul(transpose235_out1.unsqueeze::<3usize>());
        let add116_out1 = add115_out1.add(matmul195_out1);
        let layernormalization95_out1 = {
            let dtype = add116_out1.clone().dtype();
            self.layernormalization95
                .forward(add116_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear148_out1 = self.linear148.forward(layernormalization95_out1);
        let sigmoid64_out1 = burn::tensor::activation::sigmoid(linear148_out1.clone());
        let mul130_out1 = linear148_out1.mul(sigmoid64_out1);
        let linear149_out1 = self.linear149.forward(mul130_out1);
        let mul131_out1 = linear149_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add117_out1 = add116_out1.add(mul131_out1);
        let layernormalization96_out1 = {
            let dtype = add117_out1.dtype();
            self.layernormalization96
                .forward(add117_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather33_out1 = {
            let sliced = transpose1_out1.slice(s![16, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather34_out1 = {
            let sliced = transpose2_out1.slice(s![16, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization97_out1 = {
            let dtype = layernormalization96_out1.clone().dtype();
            self.layernormalization97
                .forward(
                    layernormalization96_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear150_out1 = self.linear150.forward(layernormalization97_out1);
        let sigmoid65_out1 = burn::tensor::activation::sigmoid(linear150_out1.clone());
        let mul132_out1 = linear150_out1.mul(sigmoid65_out1);
        let linear151_out1 = self.linear151.forward(mul132_out1);
        let mul133_out1 = linear151_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add118_out1 = layernormalization96_out1.add(mul133_out1);
        let layernormalization98_out1 = {
            let dtype = add118_out1.clone().dtype();
            self.layernormalization98
                .forward(add118_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat64_out1 = burn::tensor::Tensor::cat(
            [gather33_out1.clone(), layernormalization98_out1.clone()].into(),
            1,
        );
        let transpose236_out1 = gather33_out1.permute([1, 0, 2]);
        let slice71_out1 = transpose236_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose237_out1 = slice71_out1.permute([1, 0, 2]);
        let concat65_out1 = burn::tensor::Tensor::cat(
            [transpose237_out1, layernormalization98_out1.clone()].into(),
            1,
        );
        let linear152_out1 = self.linear152.forward(layernormalization98_out1);
        let reshape114_out1 = linear152_out1.reshape([1, -1, 8, 128]);
        let linear153_out1 = self.linear153.forward(concat64_out1.clone());
        let reshape115_out1 = linear153_out1.reshape([1, -1, 8, 128]);
        let linear154_out1 = self.linear154.forward(concat64_out1);
        let reshape116_out1 = linear154_out1.reshape([1, -1, 8, 128]);
        let transpose238_out1 = reshape115_out1.permute([0, 2, 1, 3]);
        let transpose239_out1 = reshape116_out1.permute([0, 2, 1, 3]);
        let linear155_out1 = self.linear155.forward(transpose6_out1);
        let reshape117_out1 = linear155_out1.reshape([1, -1, 8, 128]);
        let constant295_out1 = self.constant295.val();
        let add119_out1 = reshape114_out1
            .clone()
            .add((constant295_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose240_out1 = add119_out1.permute([0, 2, 1, 3]);
        let constant296_out1 = self.constant296.val();
        let add120_out1 = reshape114_out1
            .add((constant296_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose241_out1 = add120_out1.permute([0, 2, 1, 3]);
        let transpose242_out1 = reshape117_out1.permute([0, 2, 3, 1]);
        let matmul204_out1 = transpose241_out1.matmul(transpose242_out1);
        let pad33_out1 = matmul204_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape118_out1 = pad33_out1.reshape(concat16_out1);
        let slice72_out1 = reshape118_out1.slice(s![.., .., 1.., ..]);
        let reshape119_out1 = slice72_out1.reshape(concat17_out1);
        let transpose243_out1 = reshape119_out1.permute([3, 0, 1, 2]);
        let slice73_out1 = transpose243_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose244_out1 = slice73_out1.permute([1, 2, 3, 0]);
        let mul134_out1 = transpose244_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where49_out1 = mul134_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention17_out1,) = {
            let q = transpose240_out1;
            let k = transpose238_out1;
            let v = transpose239_out1;
            let score_bias = where49_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention17_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention17_out1 = {
                let dims = attention17_out1.dims();
                attention17_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention17_out1,)
        };
        let where50_out1 = attention17_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose245_out1 = where50_out1.permute([0, 2, 1, 3]);
        let reshape120_out1 = transpose245_out1.reshape([1, -1, 1024]);
        let linear156_out1 = self.linear156.forward(reshape120_out1);
        let add121_out1 = add118_out1.add(linear156_out1);
        (add121_out1, gather34_out1, concat65_out1, slice70_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule20 {
    layernormalization99: LayerNorm,
    squeeze34: burn::module::Param<Tensor<2>>,
    conv1d17: Conv1d,
    layernormalization100: LayerNorm,
    squeeze35: burn::module::Param<Tensor<2>>,
    layernormalization101: LayerNorm,
    linear157: Linear,
    linear158: Linear,
    layernormalization102: LayerNorm,
    layernormalization103: LayerNorm,
    linear159: Linear,
    linear160: Linear,
    layernormalization104: LayerNorm,
    linear161: Linear,
    linear162: Linear,
    linear163: Linear,
    linear164: Linear,
    constant312: burn::module::Param<Tensor<2>>,
    constant313: burn::module::Param<Tensor<2>>,
    linear165: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule20 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization99 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze34: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d17 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization100 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze35: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization101 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear157 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear158 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization102 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization103 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear159 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear160 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization104 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear161 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear162 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear163 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear164 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant312: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant313: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear165 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization99,
            squeeze34,
            conv1d17,
            layernormalization100,
            squeeze35,
            layernormalization101,
            linear157,
            linear158,
            layernormalization102,
            layernormalization103,
            linear159,
            linear160,
            layernormalization104,
            linear161,
            linear162,
            linear163,
            linear164,
            constant312,
            constant313,
            linear165,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add121_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather34_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization99_out1 = {
            let dtype = add121_out1.clone().dtype();
            self.layernormalization99
                .forward(add121_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze34_out1 = self.squeeze34.val();
        let transpose246_out1 = squeeze34_out1.permute([1, 0]);
        let matmul206_out1 = layernormalization99_out1
            .matmul(transpose246_out1.unsqueeze::<3usize>());
        let split_tensors = matmul206_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split17_out1, split17_out2] = split_tensors.try_into().unwrap();
        let sigmoid66_out1 = burn::tensor::activation::sigmoid(split17_out2);
        let mul135_out1 = split17_out1.mul(sigmoid66_out1);
        let transpose247_out1 = mul135_out1.permute([0, 2, 1]);
        let where51_out1 = transpose247_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad34_out1 = where51_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat66_out1 = burn::tensor::Tensor::cat(
            [gather34_out1, pad34_out1].into(),
            2,
        );
        let slice74_out1 = concat66_out1.clone().slice(s![.., .., - 8..]);
        let conv1d17_out1 = self.conv1d17.forward(concat66_out1);
        let transpose248_out1 = conv1d17_out1.permute([0, 2, 1]);
        let layernormalization100_out1 = {
            let dtype = transpose248_out1.dtype();
            self.layernormalization100
                .forward(transpose248_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid67_out1 = burn::tensor::activation::sigmoid(
            layernormalization100_out1.clone(),
        );
        let mul136_out1 = layernormalization100_out1.mul(sigmoid67_out1);
        let squeeze35_out1 = self.squeeze35.val();
        let transpose249_out1 = squeeze35_out1.permute([1, 0]);
        let matmul207_out1 = mul136_out1.matmul(transpose249_out1.unsqueeze::<3usize>());
        let add122_out1 = add121_out1.add(matmul207_out1);
        let layernormalization101_out1 = {
            let dtype = add122_out1.clone().dtype();
            self.layernormalization101
                .forward(add122_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear157_out1 = self.linear157.forward(layernormalization101_out1);
        let sigmoid68_out1 = burn::tensor::activation::sigmoid(linear157_out1.clone());
        let mul137_out1 = linear157_out1.mul(sigmoid68_out1);
        let linear158_out1 = self.linear158.forward(mul137_out1);
        let mul138_out1 = linear158_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add123_out1 = add122_out1.add(mul138_out1);
        let layernormalization102_out1 = {
            let dtype = add123_out1.dtype();
            self.layernormalization102
                .forward(add123_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather35_out1 = {
            let sliced = transpose1_out1.slice(s![17, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather36_out1 = {
            let sliced = transpose2_out1.slice(s![17, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization103_out1 = {
            let dtype = layernormalization102_out1.clone().dtype();
            self.layernormalization103
                .forward(
                    layernormalization102_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear159_out1 = self.linear159.forward(layernormalization103_out1);
        let sigmoid69_out1 = burn::tensor::activation::sigmoid(linear159_out1.clone());
        let mul139_out1 = linear159_out1.mul(sigmoid69_out1);
        let linear160_out1 = self.linear160.forward(mul139_out1);
        let mul140_out1 = linear160_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add124_out1 = layernormalization102_out1.add(mul140_out1);
        let layernormalization104_out1 = {
            let dtype = add124_out1.clone().dtype();
            self.layernormalization104
                .forward(add124_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat67_out1 = burn::tensor::Tensor::cat(
            [gather35_out1.clone(), layernormalization104_out1.clone()].into(),
            1,
        );
        let transpose250_out1 = gather35_out1.permute([1, 0, 2]);
        let slice75_out1 = transpose250_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose251_out1 = slice75_out1.permute([1, 0, 2]);
        let concat68_out1 = burn::tensor::Tensor::cat(
            [transpose251_out1, layernormalization104_out1.clone()].into(),
            1,
        );
        let linear161_out1 = self.linear161.forward(layernormalization104_out1);
        let reshape121_out1 = linear161_out1.reshape([1, -1, 8, 128]);
        let linear162_out1 = self.linear162.forward(concat67_out1.clone());
        let reshape122_out1 = linear162_out1.reshape([1, -1, 8, 128]);
        let linear163_out1 = self.linear163.forward(concat67_out1);
        let reshape123_out1 = linear163_out1.reshape([1, -1, 8, 128]);
        let transpose252_out1 = reshape122_out1.permute([0, 2, 1, 3]);
        let transpose253_out1 = reshape123_out1.permute([0, 2, 1, 3]);
        let linear164_out1 = self.linear164.forward(transpose6_out1);
        let reshape124_out1 = linear164_out1.reshape([1, -1, 8, 128]);
        let constant312_out1 = self.constant312.val();
        let add125_out1 = reshape121_out1
            .clone()
            .add((constant312_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose254_out1 = add125_out1.permute([0, 2, 1, 3]);
        let constant313_out1 = self.constant313.val();
        let add126_out1 = reshape121_out1
            .add((constant313_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose255_out1 = add126_out1.permute([0, 2, 1, 3]);
        let transpose256_out1 = reshape124_out1.permute([0, 2, 3, 1]);
        let matmul216_out1 = transpose255_out1.matmul(transpose256_out1);
        let pad35_out1 = matmul216_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape125_out1 = pad35_out1.reshape(concat16_out1);
        let slice76_out1 = reshape125_out1.slice(s![.., .., 1.., ..]);
        let reshape126_out1 = slice76_out1.reshape(concat17_out1);
        let transpose257_out1 = reshape126_out1.permute([3, 0, 1, 2]);
        let slice77_out1 = transpose257_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose258_out1 = slice77_out1.permute([1, 2, 3, 0]);
        let mul141_out1 = transpose258_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where52_out1 = mul141_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention18_out1,) = {
            let q = transpose254_out1;
            let k = transpose252_out1;
            let v = transpose253_out1;
            let score_bias = where52_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention18_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention18_out1 = {
                let dims = attention18_out1.dims();
                attention18_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention18_out1,)
        };
        let where53_out1 = attention18_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose259_out1 = where53_out1.permute([0, 2, 1, 3]);
        let reshape127_out1 = transpose259_out1.reshape([1, -1, 1024]);
        let linear165_out1 = self.linear165.forward(reshape127_out1);
        let add127_out1 = add124_out1.add(linear165_out1);
        (add127_out1, gather36_out1, concat68_out1, slice74_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule21 {
    layernormalization105: LayerNorm,
    squeeze36: burn::module::Param<Tensor<2>>,
    conv1d18: Conv1d,
    layernormalization106: LayerNorm,
    squeeze37: burn::module::Param<Tensor<2>>,
    layernormalization107: LayerNorm,
    linear166: Linear,
    linear167: Linear,
    layernormalization108: LayerNorm,
    layernormalization109: LayerNorm,
    linear168: Linear,
    linear169: Linear,
    layernormalization110: LayerNorm,
    linear170: Linear,
    linear171: Linear,
    linear172: Linear,
    linear173: Linear,
    constant329: burn::module::Param<Tensor<2>>,
    constant330: burn::module::Param<Tensor<2>>,
    linear174: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule21 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization105 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze36: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d18 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization106 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze37: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization107 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear166 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear167 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization108 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization109 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear168 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear169 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization110 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear170 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear171 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear172 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear173 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant329: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant330: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear174 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization105,
            squeeze36,
            conv1d18,
            layernormalization106,
            squeeze37,
            layernormalization107,
            linear166,
            linear167,
            layernormalization108,
            layernormalization109,
            linear168,
            linear169,
            layernormalization110,
            linear170,
            linear171,
            linear172,
            linear173,
            constant329,
            constant330,
            linear174,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add127_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather36_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization105_out1 = {
            let dtype = add127_out1.clone().dtype();
            self.layernormalization105
                .forward(add127_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze36_out1 = self.squeeze36.val();
        let transpose260_out1 = squeeze36_out1.permute([1, 0]);
        let matmul218_out1 = layernormalization105_out1
            .matmul(transpose260_out1.unsqueeze::<3usize>());
        let split_tensors = matmul218_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split18_out1, split18_out2] = split_tensors.try_into().unwrap();
        let sigmoid70_out1 = burn::tensor::activation::sigmoid(split18_out2);
        let mul142_out1 = split18_out1.mul(sigmoid70_out1);
        let transpose261_out1 = mul142_out1.permute([0, 2, 1]);
        let where54_out1 = transpose261_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad36_out1 = where54_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat69_out1 = burn::tensor::Tensor::cat(
            [gather36_out1, pad36_out1].into(),
            2,
        );
        let slice78_out1 = concat69_out1.clone().slice(s![.., .., - 8..]);
        let conv1d18_out1 = self.conv1d18.forward(concat69_out1);
        let transpose262_out1 = conv1d18_out1.permute([0, 2, 1]);
        let layernormalization106_out1 = {
            let dtype = transpose262_out1.dtype();
            self.layernormalization106
                .forward(transpose262_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid71_out1 = burn::tensor::activation::sigmoid(
            layernormalization106_out1.clone(),
        );
        let mul143_out1 = layernormalization106_out1.mul(sigmoid71_out1);
        let squeeze37_out1 = self.squeeze37.val();
        let transpose263_out1 = squeeze37_out1.permute([1, 0]);
        let matmul219_out1 = mul143_out1.matmul(transpose263_out1.unsqueeze::<3usize>());
        let add128_out1 = add127_out1.add(matmul219_out1);
        let layernormalization107_out1 = {
            let dtype = add128_out1.clone().dtype();
            self.layernormalization107
                .forward(add128_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear166_out1 = self.linear166.forward(layernormalization107_out1);
        let sigmoid72_out1 = burn::tensor::activation::sigmoid(linear166_out1.clone());
        let mul144_out1 = linear166_out1.mul(sigmoid72_out1);
        let linear167_out1 = self.linear167.forward(mul144_out1);
        let mul145_out1 = linear167_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add129_out1 = add128_out1.add(mul145_out1);
        let layernormalization108_out1 = {
            let dtype = add129_out1.dtype();
            self.layernormalization108
                .forward(add129_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather37_out1 = {
            let sliced = transpose1_out1.slice(s![18, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather38_out1 = {
            let sliced = transpose2_out1.slice(s![18, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization109_out1 = {
            let dtype = layernormalization108_out1.clone().dtype();
            self.layernormalization109
                .forward(
                    layernormalization108_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear168_out1 = self.linear168.forward(layernormalization109_out1);
        let sigmoid73_out1 = burn::tensor::activation::sigmoid(linear168_out1.clone());
        let mul146_out1 = linear168_out1.mul(sigmoid73_out1);
        let linear169_out1 = self.linear169.forward(mul146_out1);
        let mul147_out1 = linear169_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add130_out1 = layernormalization108_out1.add(mul147_out1);
        let layernormalization110_out1 = {
            let dtype = add130_out1.clone().dtype();
            self.layernormalization110
                .forward(add130_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat70_out1 = burn::tensor::Tensor::cat(
            [gather37_out1.clone(), layernormalization110_out1.clone()].into(),
            1,
        );
        let transpose264_out1 = gather37_out1.permute([1, 0, 2]);
        let slice79_out1 = transpose264_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose265_out1 = slice79_out1.permute([1, 0, 2]);
        let concat71_out1 = burn::tensor::Tensor::cat(
            [transpose265_out1, layernormalization110_out1.clone()].into(),
            1,
        );
        let linear170_out1 = self.linear170.forward(layernormalization110_out1);
        let reshape128_out1 = linear170_out1.reshape([1, -1, 8, 128]);
        let linear171_out1 = self.linear171.forward(concat70_out1.clone());
        let reshape129_out1 = linear171_out1.reshape([1, -1, 8, 128]);
        let linear172_out1 = self.linear172.forward(concat70_out1);
        let reshape130_out1 = linear172_out1.reshape([1, -1, 8, 128]);
        let transpose266_out1 = reshape129_out1.permute([0, 2, 1, 3]);
        let transpose267_out1 = reshape130_out1.permute([0, 2, 1, 3]);
        let linear173_out1 = self.linear173.forward(transpose6_out1);
        let reshape131_out1 = linear173_out1.reshape([1, -1, 8, 128]);
        let constant329_out1 = self.constant329.val();
        let add131_out1 = reshape128_out1
            .clone()
            .add((constant329_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose268_out1 = add131_out1.permute([0, 2, 1, 3]);
        let constant330_out1 = self.constant330.val();
        let add132_out1 = reshape128_out1
            .add((constant330_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose269_out1 = add132_out1.permute([0, 2, 1, 3]);
        let transpose270_out1 = reshape131_out1.permute([0, 2, 3, 1]);
        let matmul228_out1 = transpose269_out1.matmul(transpose270_out1);
        let pad37_out1 = matmul228_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape132_out1 = pad37_out1.reshape(concat16_out1);
        let slice80_out1 = reshape132_out1.slice(s![.., .., 1.., ..]);
        let reshape133_out1 = slice80_out1.reshape(concat17_out1);
        let transpose271_out1 = reshape133_out1.permute([3, 0, 1, 2]);
        let slice81_out1 = transpose271_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose272_out1 = slice81_out1.permute([1, 2, 3, 0]);
        let mul148_out1 = transpose272_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where55_out1 = mul148_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention19_out1,) = {
            let q = transpose268_out1;
            let k = transpose266_out1;
            let v = transpose267_out1;
            let score_bias = where55_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention19_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention19_out1 = {
                let dims = attention19_out1.dims();
                attention19_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention19_out1,)
        };
        let where56_out1 = attention19_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose273_out1 = where56_out1.permute([0, 2, 1, 3]);
        let reshape134_out1 = transpose273_out1.reshape([1, -1, 1024]);
        let linear174_out1 = self.linear174.forward(reshape134_out1);
        let add133_out1 = add130_out1.add(linear174_out1);
        (add133_out1, gather38_out1, concat71_out1, slice78_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule22 {
    layernormalization111: LayerNorm,
    squeeze38: burn::module::Param<Tensor<2>>,
    conv1d19: Conv1d,
    layernormalization112: LayerNorm,
    squeeze39: burn::module::Param<Tensor<2>>,
    layernormalization113: LayerNorm,
    linear175: Linear,
    linear176: Linear,
    layernormalization114: LayerNorm,
    layernormalization115: LayerNorm,
    linear177: Linear,
    linear178: Linear,
    layernormalization116: LayerNorm,
    linear179: Linear,
    linear180: Linear,
    linear181: Linear,
    linear182: Linear,
    constant346: burn::module::Param<Tensor<2>>,
    constant347: burn::module::Param<Tensor<2>>,
    linear183: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule22 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization111 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze38: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d19 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization112 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze39: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization113 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear175 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear176 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization114 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization115 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear177 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear178 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization116 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear179 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear180 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear181 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear182 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant346: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant347: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear183 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization111,
            squeeze38,
            conv1d19,
            layernormalization112,
            squeeze39,
            layernormalization113,
            linear175,
            linear176,
            layernormalization114,
            layernormalization115,
            linear177,
            linear178,
            layernormalization116,
            linear179,
            linear180,
            linear181,
            linear182,
            constant346,
            constant347,
            linear183,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add133_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather38_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization111_out1 = {
            let dtype = add133_out1.clone().dtype();
            self.layernormalization111
                .forward(add133_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze38_out1 = self.squeeze38.val();
        let transpose274_out1 = squeeze38_out1.permute([1, 0]);
        let matmul230_out1 = layernormalization111_out1
            .matmul(transpose274_out1.unsqueeze::<3usize>());
        let split_tensors = matmul230_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split19_out1, split19_out2] = split_tensors.try_into().unwrap();
        let sigmoid74_out1 = burn::tensor::activation::sigmoid(split19_out2);
        let mul149_out1 = split19_out1.mul(sigmoid74_out1);
        let transpose275_out1 = mul149_out1.permute([0, 2, 1]);
        let where57_out1 = transpose275_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad38_out1 = where57_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat72_out1 = burn::tensor::Tensor::cat(
            [gather38_out1, pad38_out1].into(),
            2,
        );
        let slice82_out1 = concat72_out1.clone().slice(s![.., .., - 8..]);
        let conv1d19_out1 = self.conv1d19.forward(concat72_out1);
        let transpose276_out1 = conv1d19_out1.permute([0, 2, 1]);
        let layernormalization112_out1 = {
            let dtype = transpose276_out1.dtype();
            self.layernormalization112
                .forward(transpose276_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid75_out1 = burn::tensor::activation::sigmoid(
            layernormalization112_out1.clone(),
        );
        let mul150_out1 = layernormalization112_out1.mul(sigmoid75_out1);
        let squeeze39_out1 = self.squeeze39.val();
        let transpose277_out1 = squeeze39_out1.permute([1, 0]);
        let matmul231_out1 = mul150_out1.matmul(transpose277_out1.unsqueeze::<3usize>());
        let add134_out1 = add133_out1.add(matmul231_out1);
        let layernormalization113_out1 = {
            let dtype = add134_out1.clone().dtype();
            self.layernormalization113
                .forward(add134_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear175_out1 = self.linear175.forward(layernormalization113_out1);
        let sigmoid76_out1 = burn::tensor::activation::sigmoid(linear175_out1.clone());
        let mul151_out1 = linear175_out1.mul(sigmoid76_out1);
        let linear176_out1 = self.linear176.forward(mul151_out1);
        let mul152_out1 = linear176_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add135_out1 = add134_out1.add(mul152_out1);
        let layernormalization114_out1 = {
            let dtype = add135_out1.dtype();
            self.layernormalization114
                .forward(add135_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather39_out1 = {
            let sliced = transpose1_out1.slice(s![19, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather40_out1 = {
            let sliced = transpose2_out1.slice(s![19, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization115_out1 = {
            let dtype = layernormalization114_out1.clone().dtype();
            self.layernormalization115
                .forward(
                    layernormalization114_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear177_out1 = self.linear177.forward(layernormalization115_out1);
        let sigmoid77_out1 = burn::tensor::activation::sigmoid(linear177_out1.clone());
        let mul153_out1 = linear177_out1.mul(sigmoid77_out1);
        let linear178_out1 = self.linear178.forward(mul153_out1);
        let mul154_out1 = linear178_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add136_out1 = layernormalization114_out1.add(mul154_out1);
        let layernormalization116_out1 = {
            let dtype = add136_out1.clone().dtype();
            self.layernormalization116
                .forward(add136_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat73_out1 = burn::tensor::Tensor::cat(
            [gather39_out1.clone(), layernormalization116_out1.clone()].into(),
            1,
        );
        let transpose278_out1 = gather39_out1.permute([1, 0, 2]);
        let slice83_out1 = transpose278_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose279_out1 = slice83_out1.permute([1, 0, 2]);
        let concat74_out1 = burn::tensor::Tensor::cat(
            [transpose279_out1, layernormalization116_out1.clone()].into(),
            1,
        );
        let linear179_out1 = self.linear179.forward(layernormalization116_out1);
        let reshape135_out1 = linear179_out1.reshape([1, -1, 8, 128]);
        let linear180_out1 = self.linear180.forward(concat73_out1.clone());
        let reshape136_out1 = linear180_out1.reshape([1, -1, 8, 128]);
        let linear181_out1 = self.linear181.forward(concat73_out1);
        let reshape137_out1 = linear181_out1.reshape([1, -1, 8, 128]);
        let transpose280_out1 = reshape136_out1.permute([0, 2, 1, 3]);
        let transpose281_out1 = reshape137_out1.permute([0, 2, 1, 3]);
        let linear182_out1 = self.linear182.forward(transpose6_out1);
        let reshape138_out1 = linear182_out1.reshape([1, -1, 8, 128]);
        let constant346_out1 = self.constant346.val();
        let add137_out1 = reshape135_out1
            .clone()
            .add((constant346_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose282_out1 = add137_out1.permute([0, 2, 1, 3]);
        let constant347_out1 = self.constant347.val();
        let add138_out1 = reshape135_out1
            .add((constant347_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose283_out1 = add138_out1.permute([0, 2, 1, 3]);
        let transpose284_out1 = reshape138_out1.permute([0, 2, 3, 1]);
        let matmul240_out1 = transpose283_out1.matmul(transpose284_out1);
        let pad39_out1 = matmul240_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape139_out1 = pad39_out1.reshape(concat16_out1);
        let slice84_out1 = reshape139_out1.slice(s![.., .., 1.., ..]);
        let reshape140_out1 = slice84_out1.reshape(concat17_out1);
        let transpose285_out1 = reshape140_out1.permute([3, 0, 1, 2]);
        let slice85_out1 = transpose285_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose286_out1 = slice85_out1.permute([1, 2, 3, 0]);
        let mul155_out1 = transpose286_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where58_out1 = mul155_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention20_out1,) = {
            let q = transpose282_out1;
            let k = transpose280_out1;
            let v = transpose281_out1;
            let score_bias = where58_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention20_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention20_out1 = {
                let dims = attention20_out1.dims();
                attention20_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention20_out1,)
        };
        let where59_out1 = attention20_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose287_out1 = where59_out1.permute([0, 2, 1, 3]);
        let reshape141_out1 = transpose287_out1.reshape([1, -1, 1024]);
        let linear183_out1 = self.linear183.forward(reshape141_out1);
        let add139_out1 = add136_out1.add(linear183_out1);
        (add139_out1, gather40_out1, concat74_out1, slice82_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule23 {
    layernormalization117: LayerNorm,
    squeeze40: burn::module::Param<Tensor<2>>,
    conv1d20: Conv1d,
    layernormalization118: LayerNorm,
    squeeze41: burn::module::Param<Tensor<2>>,
    layernormalization119: LayerNorm,
    linear184: Linear,
    linear185: Linear,
    layernormalization120: LayerNorm,
    layernormalization121: LayerNorm,
    linear186: Linear,
    linear187: Linear,
    layernormalization122: LayerNorm,
    linear188: Linear,
    linear189: Linear,
    linear190: Linear,
    linear191: Linear,
    constant363: burn::module::Param<Tensor<2>>,
    constant364: burn::module::Param<Tensor<2>>,
    linear192: Linear,
    #[module(skip)]
    device: Device,
}
impl Submodule23 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization117 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze40: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d20 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization118 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze41: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization119 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear184 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear185 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization120 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization121 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear186 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear187 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization122 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear188 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear189 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear190 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear191 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant363: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant364: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear192 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        Self {
            layernormalization117,
            squeeze40,
            conv1d20,
            layernormalization118,
            squeeze41,
            layernormalization119,
            linear184,
            linear185,
            layernormalization120,
            layernormalization121,
            linear186,
            linear187,
            layernormalization122,
            linear188,
            linear189,
            linear190,
            linear191,
            constant363,
            constant364,
            linear192,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add139_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather40_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (Tensor<3>, Tensor<3>, Tensor<3>, Tensor<3>) {
        let layernormalization117_out1 = {
            let dtype = add139_out1.clone().dtype();
            self.layernormalization117
                .forward(add139_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze40_out1 = self.squeeze40.val();
        let transpose288_out1 = squeeze40_out1.permute([1, 0]);
        let matmul242_out1 = layernormalization117_out1
            .matmul(transpose288_out1.unsqueeze::<3usize>());
        let split_tensors = matmul242_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split20_out1, split20_out2] = split_tensors.try_into().unwrap();
        let sigmoid78_out1 = burn::tensor::activation::sigmoid(split20_out2);
        let mul156_out1 = split20_out1.mul(sigmoid78_out1);
        let transpose289_out1 = mul156_out1.permute([0, 2, 1]);
        let where60_out1 = transpose289_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad40_out1 = where60_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat75_out1 = burn::tensor::Tensor::cat(
            [gather40_out1, pad40_out1].into(),
            2,
        );
        let slice86_out1 = concat75_out1.clone().slice(s![.., .., - 8..]);
        let conv1d20_out1 = self.conv1d20.forward(concat75_out1);
        let transpose290_out1 = conv1d20_out1.permute([0, 2, 1]);
        let layernormalization118_out1 = {
            let dtype = transpose290_out1.dtype();
            self.layernormalization118
                .forward(transpose290_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid79_out1 = burn::tensor::activation::sigmoid(
            layernormalization118_out1.clone(),
        );
        let mul157_out1 = layernormalization118_out1.mul(sigmoid79_out1);
        let squeeze41_out1 = self.squeeze41.val();
        let transpose291_out1 = squeeze41_out1.permute([1, 0]);
        let matmul243_out1 = mul157_out1.matmul(transpose291_out1.unsqueeze::<3usize>());
        let add140_out1 = add139_out1.add(matmul243_out1);
        let layernormalization119_out1 = {
            let dtype = add140_out1.clone().dtype();
            self.layernormalization119
                .forward(add140_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear184_out1 = self.linear184.forward(layernormalization119_out1);
        let sigmoid80_out1 = burn::tensor::activation::sigmoid(linear184_out1.clone());
        let mul158_out1 = linear184_out1.mul(sigmoid80_out1);
        let linear185_out1 = self.linear185.forward(mul158_out1);
        let mul159_out1 = linear185_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add141_out1 = add140_out1.add(mul159_out1);
        let layernormalization120_out1 = {
            let dtype = add141_out1.dtype();
            self.layernormalization120
                .forward(add141_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather41_out1 = {
            let sliced = transpose1_out1.slice(s![20, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather42_out1 = {
            let sliced = transpose2_out1.slice(s![20, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization121_out1 = {
            let dtype = layernormalization120_out1.clone().dtype();
            self.layernormalization121
                .forward(
                    layernormalization120_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear186_out1 = self.linear186.forward(layernormalization121_out1);
        let sigmoid81_out1 = burn::tensor::activation::sigmoid(linear186_out1.clone());
        let mul160_out1 = linear186_out1.mul(sigmoid81_out1);
        let linear187_out1 = self.linear187.forward(mul160_out1);
        let mul161_out1 = linear187_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add142_out1 = layernormalization120_out1.add(mul161_out1);
        let layernormalization122_out1 = {
            let dtype = add142_out1.clone().dtype();
            self.layernormalization122
                .forward(add142_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat76_out1 = burn::tensor::Tensor::cat(
            [gather41_out1.clone(), layernormalization122_out1.clone()].into(),
            1,
        );
        let transpose292_out1 = gather41_out1.permute([1, 0, 2]);
        let slice87_out1 = transpose292_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose293_out1 = slice87_out1.permute([1, 0, 2]);
        let concat77_out1 = burn::tensor::Tensor::cat(
            [transpose293_out1, layernormalization122_out1.clone()].into(),
            1,
        );
        let linear188_out1 = self.linear188.forward(layernormalization122_out1);
        let reshape142_out1 = linear188_out1.reshape([1, -1, 8, 128]);
        let linear189_out1 = self.linear189.forward(concat76_out1.clone());
        let reshape143_out1 = linear189_out1.reshape([1, -1, 8, 128]);
        let linear190_out1 = self.linear190.forward(concat76_out1);
        let reshape144_out1 = linear190_out1.reshape([1, -1, 8, 128]);
        let transpose294_out1 = reshape143_out1.permute([0, 2, 1, 3]);
        let transpose295_out1 = reshape144_out1.permute([0, 2, 1, 3]);
        let linear191_out1 = self.linear191.forward(transpose6_out1);
        let reshape145_out1 = linear191_out1.reshape([1, -1, 8, 128]);
        let constant363_out1 = self.constant363.val();
        let add143_out1 = reshape142_out1
            .clone()
            .add((constant363_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose296_out1 = add143_out1.permute([0, 2, 1, 3]);
        let constant364_out1 = self.constant364.val();
        let add144_out1 = reshape142_out1
            .add((constant364_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose297_out1 = add144_out1.permute([0, 2, 1, 3]);
        let transpose298_out1 = reshape145_out1.permute([0, 2, 3, 1]);
        let matmul252_out1 = transpose297_out1.matmul(transpose298_out1);
        let pad41_out1 = matmul252_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape146_out1 = pad41_out1.reshape(concat16_out1);
        let slice88_out1 = reshape146_out1.slice(s![.., .., 1.., ..]);
        let reshape147_out1 = slice88_out1.reshape(concat17_out1);
        let transpose299_out1 = reshape147_out1.permute([3, 0, 1, 2]);
        let slice89_out1 = transpose299_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose300_out1 = slice89_out1.permute([1, 2, 3, 0]);
        let mul162_out1 = transpose300_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where61_out1 = mul162_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention21_out1,) = {
            let q = transpose296_out1;
            let k = transpose294_out1;
            let v = transpose295_out1;
            let score_bias = where61_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention21_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention21_out1 = {
                let dims = attention21_out1.dims();
                attention21_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention21_out1,)
        };
        let where62_out1 = attention21_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose301_out1 = where62_out1.permute([0, 2, 1, 3]);
        let reshape148_out1 = transpose301_out1.reshape([1, -1, 1024]);
        let linear192_out1 = self.linear192.forward(reshape148_out1);
        let add145_out1 = add142_out1.add(linear192_out1);
        (add145_out1, gather42_out1, concat77_out1, slice86_out1)
    }
}
#[derive(Module, Debug)]
pub struct Submodule24 {
    layernormalization123: LayerNorm,
    squeeze42: burn::module::Param<Tensor<2>>,
    conv1d21: Conv1d,
    layernormalization124: LayerNorm,
    squeeze43: burn::module::Param<Tensor<2>>,
    layernormalization125: LayerNorm,
    linear193: Linear,
    linear194: Linear,
    layernormalization126: LayerNorm,
    layernormalization127: LayerNorm,
    linear195: Linear,
    linear196: Linear,
    layernormalization128: LayerNorm,
    linear197: Linear,
    linear198: Linear,
    linear199: Linear,
    linear200: Linear,
    constant380: burn::module::Param<Tensor<2>>,
    constant381: burn::module::Param<Tensor<2>>,
    linear201: Linear,
    layernormalization129: LayerNorm,
    squeeze44: burn::module::Param<Tensor<2>>,
    conv1d22: Conv1d,
    layernormalization130: LayerNorm,
    squeeze45: burn::module::Param<Tensor<2>>,
    layernormalization131: LayerNorm,
    linear202: Linear,
    linear203: Linear,
    layernormalization132: LayerNorm,
    layernormalization133: LayerNorm,
    linear204: Linear,
    linear205: Linear,
    layernormalization134: LayerNorm,
    linear206: Linear,
    linear207: Linear,
    linear208: Linear,
    linear209: Linear,
    constant397: burn::module::Param<Tensor<2>>,
    constant398: burn::module::Param<Tensor<2>>,
    linear210: Linear,
    layernormalization135: LayerNorm,
    squeeze46: burn::module::Param<Tensor<2>>,
    conv1d23: Conv1d,
    layernormalization136: LayerNorm,
    squeeze47: burn::module::Param<Tensor<2>>,
    layernormalization137: LayerNorm,
    linear211: Linear,
    linear212: Linear,
    layernormalization138: LayerNorm,
    layernormalization139: LayerNorm,
    linear213: Linear,
    linear214: Linear,
    layernormalization140: LayerNorm,
    linear215: Linear,
    linear216: Linear,
    linear217: Linear,
    linear218: Linear,
    constant414: burn::module::Param<Tensor<2>>,
    constant415: burn::module::Param<Tensor<2>>,
    linear219: Linear,
    layernormalization141: LayerNorm,
    squeeze48: burn::module::Param<Tensor<2>>,
    #[module(skip)]
    device: Device,
}
impl Submodule24 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let layernormalization123 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze42: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d21 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization124 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze43: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization125 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear193 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear194 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization126 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization127 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear195 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear196 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization128 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear197 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear198 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear199 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear200 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant380: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant381: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear201 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let layernormalization129 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze44: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d22 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization130 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze45: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization131 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear202 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear203 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization132 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization133 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear204 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear205 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization134 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear206 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear207 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear208 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear209 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant397: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant398: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear210 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let layernormalization135 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze46: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        let conv1d23 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization136 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze47: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization137 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear211 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear212 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization138 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let layernormalization139 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear213 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear214 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization140 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear215 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear216 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear217 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let linear218 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let constant414: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let constant415: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([8, 128], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [8, 128].into(),
        );
        let linear219 = LinearConfig::new(1024, 1024).with_bias(false).init(device);
        let layernormalization141 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze48: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([2048, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [2048, 1024].into(),
        );
        Self {
            layernormalization123,
            squeeze42,
            conv1d21,
            layernormalization124,
            squeeze43,
            layernormalization125,
            linear193,
            linear194,
            layernormalization126,
            layernormalization127,
            linear195,
            linear196,
            layernormalization128,
            linear197,
            linear198,
            linear199,
            linear200,
            constant380,
            constant381,
            linear201,
            layernormalization129,
            squeeze44,
            conv1d22,
            layernormalization130,
            squeeze45,
            layernormalization131,
            linear202,
            linear203,
            layernormalization132,
            layernormalization133,
            linear204,
            linear205,
            layernormalization134,
            linear206,
            linear207,
            linear208,
            linear209,
            constant397,
            constant398,
            linear210,
            layernormalization135,
            squeeze46,
            conv1d23,
            layernormalization136,
            squeeze47,
            layernormalization137,
            linear211,
            linear212,
            layernormalization138,
            layernormalization139,
            linear213,
            linear214,
            layernormalization140,
            linear215,
            linear216,
            linear217,
            linear218,
            constant414,
            constant415,
            linear219,
            layernormalization141,
            squeeze48,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        add145_out1: Tensor<3>,
        unsqueeze33_out1: Tensor<3, Bool>,
        constant453_out1: f32,
        gather42_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        transpose1_out1: Tensor<4>,
        transpose2_out1: Tensor<4>,
        unsqueeze29_out1: [i64; 1],
        transpose6_out1: Tensor<3>,
        concat16_out1: [i64; 4],
        concat17_out1: [i64; 4],
        unsqueeze19_out1: [i64; 1],
        constant687_out1: Tensor<1>,
        unsqueeze31_out1: Tensor<4, Bool>,
        constant450_out1: f32,
        expand15_out1: Tensor<4, Bool>,
    ) -> (
        Tensor<3>,
        Tensor<3>,
        Tensor<3>,
        Tensor<3>,
        Tensor<3>,
        Tensor<3>,
        Tensor<3>,
        Tensor<3>,
    ) {
        let layernormalization123_out1 = {
            let dtype = add145_out1.clone().dtype();
            self.layernormalization123
                .forward(add145_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze42_out1 = self.squeeze42.val();
        let transpose302_out1 = squeeze42_out1.permute([1, 0]);
        let matmul254_out1 = layernormalization123_out1
            .matmul(transpose302_out1.unsqueeze::<3usize>());
        let split_tensors = matmul254_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split21_out1, split21_out2] = split_tensors.try_into().unwrap();
        let sigmoid82_out1 = burn::tensor::activation::sigmoid(split21_out2);
        let mul163_out1 = split21_out1.mul(sigmoid82_out1);
        let transpose303_out1 = mul163_out1.permute([0, 2, 1]);
        let where63_out1 = transpose303_out1
            .mask_fill(unsqueeze33_out1.clone(), constant453_out1);
        let pad42_out1 = where63_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat78_out1 = burn::tensor::Tensor::cat(
            [gather42_out1, pad42_out1].into(),
            2,
        );
        let slice90_out1 = concat78_out1.clone().slice(s![.., .., - 8..]);
        let conv1d21_out1 = self.conv1d21.forward(concat78_out1);
        let transpose304_out1 = conv1d21_out1.permute([0, 2, 1]);
        let layernormalization124_out1 = {
            let dtype = transpose304_out1.dtype();
            self.layernormalization124
                .forward(transpose304_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid83_out1 = burn::tensor::activation::sigmoid(
            layernormalization124_out1.clone(),
        );
        let mul164_out1 = layernormalization124_out1.mul(sigmoid83_out1);
        let squeeze43_out1 = self.squeeze43.val();
        let transpose305_out1 = squeeze43_out1.permute([1, 0]);
        let matmul255_out1 = mul164_out1.matmul(transpose305_out1.unsqueeze::<3usize>());
        let add146_out1 = add145_out1.add(matmul255_out1);
        let layernormalization125_out1 = {
            let dtype = add146_out1.clone().dtype();
            self.layernormalization125
                .forward(add146_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear193_out1 = self.linear193.forward(layernormalization125_out1);
        let sigmoid84_out1 = burn::tensor::activation::sigmoid(linear193_out1.clone());
        let mul165_out1 = linear193_out1.mul(sigmoid84_out1);
        let linear194_out1 = self.linear194.forward(mul165_out1);
        let mul166_out1 = linear194_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add147_out1 = add146_out1.add(mul166_out1);
        let layernormalization126_out1 = {
            let dtype = add147_out1.dtype();
            self.layernormalization126
                .forward(add147_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather43_out1 = {
            let sliced = transpose1_out1.clone().slice(s![21, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather44_out1 = {
            let sliced = transpose2_out1.clone().slice(s![21, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization127_out1 = {
            let dtype = layernormalization126_out1.clone().dtype();
            self.layernormalization127
                .forward(
                    layernormalization126_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear195_out1 = self.linear195.forward(layernormalization127_out1);
        let sigmoid85_out1 = burn::tensor::activation::sigmoid(linear195_out1.clone());
        let mul167_out1 = linear195_out1.mul(sigmoid85_out1);
        let linear196_out1 = self.linear196.forward(mul167_out1);
        let mul168_out1 = linear196_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add148_out1 = layernormalization126_out1.add(mul168_out1);
        let layernormalization128_out1 = {
            let dtype = add148_out1.clone().dtype();
            self.layernormalization128
                .forward(add148_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat79_out1 = burn::tensor::Tensor::cat(
            [gather43_out1.clone(), layernormalization128_out1.clone()].into(),
            1,
        );
        let transpose306_out1 = gather43_out1.permute([1, 0, 2]);
        let slice91_out1 = transpose306_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose307_out1 = slice91_out1.permute([1, 0, 2]);
        let concat80_out1 = burn::tensor::Tensor::cat(
            [transpose307_out1, layernormalization128_out1.clone()].into(),
            1,
        );
        let linear197_out1 = self.linear197.forward(layernormalization128_out1);
        let reshape149_out1 = linear197_out1.reshape([1, -1, 8, 128]);
        let linear198_out1 = self.linear198.forward(concat79_out1.clone());
        let reshape150_out1 = linear198_out1.reshape([1, -1, 8, 128]);
        let linear199_out1 = self.linear199.forward(concat79_out1);
        let reshape151_out1 = linear199_out1.reshape([1, -1, 8, 128]);
        let transpose308_out1 = reshape150_out1.permute([0, 2, 1, 3]);
        let transpose309_out1 = reshape151_out1.permute([0, 2, 1, 3]);
        let linear200_out1 = self.linear200.forward(transpose6_out1.clone());
        let reshape152_out1 = linear200_out1.reshape([1, -1, 8, 128]);
        let constant380_out1 = self.constant380.val();
        let add149_out1 = reshape149_out1
            .clone()
            .add((constant380_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose310_out1 = add149_out1.permute([0, 2, 1, 3]);
        let constant381_out1 = self.constant381.val();
        let add150_out1 = reshape149_out1
            .add((constant381_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose311_out1 = add150_out1.permute([0, 2, 1, 3]);
        let transpose312_out1 = reshape152_out1.permute([0, 2, 3, 1]);
        let matmul264_out1 = transpose311_out1.matmul(transpose312_out1);
        let pad43_out1 = matmul264_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape153_out1 = pad43_out1.reshape(concat16_out1);
        let slice92_out1 = reshape153_out1.slice(s![.., .., 1.., ..]);
        let reshape154_out1 = slice92_out1.reshape(concat17_out1);
        let transpose313_out1 = reshape154_out1.permute([3, 0, 1, 2]);
        let slice93_out1 = transpose313_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose314_out1 = slice93_out1.permute([1, 2, 3, 0]);
        let mul169_out1 = transpose314_out1
            .mul((constant687_out1.clone()).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where64_out1 = mul169_out1
            .mask_fill(unsqueeze31_out1.clone(), constant450_out1);
        let (attention22_out1,) = {
            let q = transpose310_out1;
            let k = transpose308_out1;
            let v = transpose309_out1;
            let score_bias = where64_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention22_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention22_out1 = {
                let dims = attention22_out1.dims();
                attention22_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention22_out1,)
        };
        let where65_out1 = attention22_out1
            .mask_fill(expand15_out1.clone(), constant453_out1);
        let transpose315_out1 = where65_out1.permute([0, 2, 1, 3]);
        let reshape155_out1 = transpose315_out1.reshape([1, -1, 1024]);
        let linear201_out1 = self.linear201.forward(reshape155_out1);
        let add151_out1 = add148_out1.add(linear201_out1);
        let layernormalization129_out1 = {
            let dtype = add151_out1.clone().dtype();
            self.layernormalization129
                .forward(add151_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze44_out1 = self.squeeze44.val();
        let transpose316_out1 = squeeze44_out1.permute([1, 0]);
        let matmul266_out1 = layernormalization129_out1
            .matmul(transpose316_out1.unsqueeze::<3usize>());
        let split_tensors = matmul266_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split22_out1, split22_out2] = split_tensors.try_into().unwrap();
        let sigmoid86_out1 = burn::tensor::activation::sigmoid(split22_out2);
        let mul170_out1 = split22_out1.mul(sigmoid86_out1);
        let transpose317_out1 = mul170_out1.permute([0, 2, 1]);
        let where66_out1 = transpose317_out1
            .mask_fill(unsqueeze33_out1.clone(), constant453_out1);
        let pad44_out1 = where66_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat81_out1 = burn::tensor::Tensor::cat(
            [gather44_out1, pad44_out1].into(),
            2,
        );
        let slice94_out1 = concat81_out1.clone().slice(s![.., .., - 8..]);
        let conv1d22_out1 = self.conv1d22.forward(concat81_out1);
        let transpose318_out1 = conv1d22_out1.permute([0, 2, 1]);
        let layernormalization130_out1 = {
            let dtype = transpose318_out1.dtype();
            self.layernormalization130
                .forward(transpose318_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid87_out1 = burn::tensor::activation::sigmoid(
            layernormalization130_out1.clone(),
        );
        let mul171_out1 = layernormalization130_out1.mul(sigmoid87_out1);
        let squeeze45_out1 = self.squeeze45.val();
        let transpose319_out1 = squeeze45_out1.permute([1, 0]);
        let matmul267_out1 = mul171_out1.matmul(transpose319_out1.unsqueeze::<3usize>());
        let add152_out1 = add151_out1.add(matmul267_out1);
        let layernormalization131_out1 = {
            let dtype = add152_out1.clone().dtype();
            self.layernormalization131
                .forward(add152_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear202_out1 = self.linear202.forward(layernormalization131_out1);
        let sigmoid88_out1 = burn::tensor::activation::sigmoid(linear202_out1.clone());
        let mul172_out1 = linear202_out1.mul(sigmoid88_out1);
        let linear203_out1 = self.linear203.forward(mul172_out1);
        let mul173_out1 = linear203_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add153_out1 = add152_out1.add(mul173_out1);
        let layernormalization132_out1 = {
            let dtype = add153_out1.dtype();
            self.layernormalization132
                .forward(add153_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather45_out1 = {
            let sliced = transpose1_out1.clone().slice(s![22, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather46_out1 = {
            let sliced = transpose2_out1.clone().slice(s![22, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization133_out1 = {
            let dtype = layernormalization132_out1.clone().dtype();
            self.layernormalization133
                .forward(
                    layernormalization132_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear204_out1 = self.linear204.forward(layernormalization133_out1);
        let sigmoid89_out1 = burn::tensor::activation::sigmoid(linear204_out1.clone());
        let mul174_out1 = linear204_out1.mul(sigmoid89_out1);
        let linear205_out1 = self.linear205.forward(mul174_out1);
        let mul175_out1 = linear205_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add154_out1 = layernormalization132_out1.add(mul175_out1);
        let layernormalization134_out1 = {
            let dtype = add154_out1.clone().dtype();
            self.layernormalization134
                .forward(add154_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat82_out1 = burn::tensor::Tensor::cat(
            [gather45_out1.clone(), layernormalization134_out1.clone()].into(),
            1,
        );
        let transpose320_out1 = gather45_out1.permute([1, 0, 2]);
        let slice95_out1 = transpose320_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose321_out1 = slice95_out1.permute([1, 0, 2]);
        let concat83_out1 = burn::tensor::Tensor::cat(
            [transpose321_out1, layernormalization134_out1.clone()].into(),
            1,
        );
        let linear206_out1 = self.linear206.forward(layernormalization134_out1);
        let reshape156_out1 = linear206_out1.reshape([1, -1, 8, 128]);
        let linear207_out1 = self.linear207.forward(concat82_out1.clone());
        let reshape157_out1 = linear207_out1.reshape([1, -1, 8, 128]);
        let linear208_out1 = self.linear208.forward(concat82_out1);
        let reshape158_out1 = linear208_out1.reshape([1, -1, 8, 128]);
        let transpose322_out1 = reshape157_out1.permute([0, 2, 1, 3]);
        let transpose323_out1 = reshape158_out1.permute([0, 2, 1, 3]);
        let linear209_out1 = self.linear209.forward(transpose6_out1.clone());
        let reshape159_out1 = linear209_out1.reshape([1, -1, 8, 128]);
        let constant397_out1 = self.constant397.val();
        let add155_out1 = reshape156_out1
            .clone()
            .add((constant397_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose324_out1 = add155_out1.permute([0, 2, 1, 3]);
        let constant398_out1 = self.constant398.val();
        let add156_out1 = reshape156_out1
            .add((constant398_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose325_out1 = add156_out1.permute([0, 2, 1, 3]);
        let transpose326_out1 = reshape159_out1.permute([0, 2, 3, 1]);
        let matmul276_out1 = transpose325_out1.matmul(transpose326_out1);
        let pad45_out1 = matmul276_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape160_out1 = pad45_out1.reshape(concat16_out1);
        let slice96_out1 = reshape160_out1.slice(s![.., .., 1.., ..]);
        let reshape161_out1 = slice96_out1.reshape(concat17_out1);
        let transpose327_out1 = reshape161_out1.permute([3, 0, 1, 2]);
        let slice97_out1 = transpose327_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose328_out1 = slice97_out1.permute([1, 2, 3, 0]);
        let mul176_out1 = transpose328_out1
            .mul((constant687_out1.clone()).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where67_out1 = mul176_out1
            .mask_fill(unsqueeze31_out1.clone(), constant450_out1);
        let (attention23_out1,) = {
            let q = transpose324_out1;
            let k = transpose322_out1;
            let v = transpose323_out1;
            let score_bias = where67_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention23_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention23_out1 = {
                let dims = attention23_out1.dims();
                attention23_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention23_out1,)
        };
        let where68_out1 = attention23_out1
            .mask_fill(expand15_out1.clone(), constant453_out1);
        let transpose329_out1 = where68_out1.permute([0, 2, 1, 3]);
        let reshape162_out1 = transpose329_out1.reshape([1, -1, 1024]);
        let linear210_out1 = self.linear210.forward(reshape162_out1);
        let add157_out1 = add154_out1.add(linear210_out1);
        let layernormalization135_out1 = {
            let dtype = add157_out1.clone().dtype();
            self.layernormalization135
                .forward(add157_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze46_out1 = self.squeeze46.val();
        let transpose330_out1 = squeeze46_out1.permute([1, 0]);
        let matmul278_out1 = layernormalization135_out1
            .matmul(transpose330_out1.unsqueeze::<3usize>());
        let split_tensors = matmul278_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split23_out1, split23_out2] = split_tensors.try_into().unwrap();
        let sigmoid90_out1 = burn::tensor::activation::sigmoid(split23_out2);
        let mul177_out1 = split23_out1.mul(sigmoid90_out1);
        let transpose331_out1 = mul177_out1.permute([0, 2, 1]);
        let where69_out1 = transpose331_out1
            .mask_fill(unsqueeze33_out1.clone(), constant453_out1);
        let pad46_out1 = where69_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat84_out1 = burn::tensor::Tensor::cat(
            [gather46_out1, pad46_out1].into(),
            2,
        );
        let slice98_out1 = concat84_out1.clone().slice(s![.., .., - 8..]);
        let conv1d23_out1 = self.conv1d23.forward(concat84_out1);
        let transpose332_out1 = conv1d23_out1.permute([0, 2, 1]);
        let layernormalization136_out1 = {
            let dtype = transpose332_out1.dtype();
            self.layernormalization136
                .forward(transpose332_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid91_out1 = burn::tensor::activation::sigmoid(
            layernormalization136_out1.clone(),
        );
        let mul178_out1 = layernormalization136_out1.mul(sigmoid91_out1);
        let squeeze47_out1 = self.squeeze47.val();
        let transpose333_out1 = squeeze47_out1.permute([1, 0]);
        let matmul279_out1 = mul178_out1.matmul(transpose333_out1.unsqueeze::<3usize>());
        let add158_out1 = add157_out1.add(matmul279_out1);
        let layernormalization137_out1 = {
            let dtype = add158_out1.clone().dtype();
            self.layernormalization137
                .forward(add158_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear211_out1 = self.linear211.forward(layernormalization137_out1);
        let sigmoid92_out1 = burn::tensor::activation::sigmoid(linear211_out1.clone());
        let mul179_out1 = linear211_out1.mul(sigmoid92_out1);
        let linear212_out1 = self.linear212.forward(mul179_out1);
        let mul180_out1 = linear212_out1
            .mul((constant684_out1.clone()).unsqueeze_dims(&[0isize, 1isize]));
        let add159_out1 = add158_out1.add(mul180_out1);
        let layernormalization138_out1 = {
            let dtype = add159_out1.dtype();
            self.layernormalization138
                .forward(add159_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let gather47_out1 = {
            let sliced = transpose1_out1.slice(s![23, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let gather48_out1 = {
            let sliced = transpose2_out1.slice(s![23, .., .., ..]);
            sliced.squeeze_dim::<3usize>(0)
        };
        let layernormalization139_out1 = {
            let dtype = layernormalization138_out1.clone().dtype();
            self.layernormalization139
                .forward(
                    layernormalization138_out1.clone().cast(burn::tensor::DType::F32),
                )
                .cast(dtype)
        };
        let linear213_out1 = self.linear213.forward(layernormalization139_out1);
        let sigmoid93_out1 = burn::tensor::activation::sigmoid(linear213_out1.clone());
        let mul181_out1 = linear213_out1.mul(sigmoid93_out1);
        let linear214_out1 = self.linear214.forward(mul181_out1);
        let mul182_out1 = linear214_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add160_out1 = layernormalization138_out1.add(mul182_out1);
        let layernormalization140_out1 = {
            let dtype = add160_out1.clone().dtype();
            self.layernormalization140
                .forward(add160_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let concat85_out1 = burn::tensor::Tensor::cat(
            [gather47_out1.clone(), layernormalization140_out1.clone()].into(),
            1,
        );
        let transpose334_out1 = gather47_out1.permute([1, 0, 2]);
        let slice99_out1 = transpose334_out1
            .slice(s![unsqueeze29_out1[0]..9223372036854775807, .., ..]);
        let transpose335_out1 = slice99_out1.permute([1, 0, 2]);
        let concat86_out1 = burn::tensor::Tensor::cat(
            [transpose335_out1, layernormalization140_out1.clone()].into(),
            1,
        );
        let linear215_out1 = self.linear215.forward(layernormalization140_out1);
        let reshape163_out1 = linear215_out1.reshape([1, -1, 8, 128]);
        let linear216_out1 = self.linear216.forward(concat85_out1.clone());
        let reshape164_out1 = linear216_out1.reshape([1, -1, 8, 128]);
        let linear217_out1 = self.linear217.forward(concat85_out1);
        let reshape165_out1 = linear217_out1.reshape([1, -1, 8, 128]);
        let transpose336_out1 = reshape164_out1.permute([0, 2, 1, 3]);
        let transpose337_out1 = reshape165_out1.permute([0, 2, 1, 3]);
        let linear218_out1 = self.linear218.forward(transpose6_out1);
        let reshape166_out1 = linear218_out1.reshape([1, -1, 8, 128]);
        let constant414_out1 = self.constant414.val();
        let add161_out1 = reshape163_out1
            .clone()
            .add((constant414_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose338_out1 = add161_out1.permute([0, 2, 1, 3]);
        let constant415_out1 = self.constant415.val();
        let add162_out1 = reshape163_out1
            .add((constant415_out1).unsqueeze_dims(&[0isize, 1isize]));
        let transpose339_out1 = add162_out1.permute([0, 2, 1, 3]);
        let transpose340_out1 = reshape166_out1.permute([0, 2, 3, 1]);
        let matmul288_out1 = transpose339_out1.matmul(transpose340_out1);
        let pad47_out1 = matmul288_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize), (1usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let reshape167_out1 = pad47_out1.reshape(concat16_out1);
        let slice100_out1 = reshape167_out1.slice(s![.., .., 1.., ..]);
        let reshape168_out1 = slice100_out1.reshape(concat17_out1);
        let transpose341_out1 = reshape168_out1.permute([3, 0, 1, 2]);
        let slice101_out1 = transpose341_out1
            .slice(s![0..unsqueeze19_out1[0], .., .., ..]);
        let transpose342_out1 = slice101_out1.permute([1, 2, 3, 0]);
        let mul183_out1 = transpose342_out1
            .mul((constant687_out1).unsqueeze_dims(&[0isize, 1isize, 2isize]));
        let where70_out1 = mul183_out1.mask_fill(unsqueeze31_out1, constant450_out1);
        let (attention24_out1,) = {
            let q = transpose338_out1;
            let k = transpose336_out1;
            let v = transpose337_out1;
            let score_bias = where70_out1;
            let hidden_rows = score_bias
                .clone()
                .max_dim(3)
                .equal_elem(f32::NEG_INFINITY);
            let attention24_out1 = burn::tensor::module::attention(
                q,
                k,
                v,
                None,
                Some(score_bias),
                burn::tensor::ops::AttentionModuleOptions {
                    scale: None,
                    softcap: None,
                    is_causal: false,
                },
            );
            let attention24_out1 = {
                let dims = attention24_out1.dims();
                attention24_out1.mask_fill(hidden_rows.expand(dims), 0.0)
            };
            (attention24_out1,)
        };
        let where71_out1 = attention24_out1.mask_fill(expand15_out1, constant453_out1);
        let transpose343_out1 = where71_out1.permute([0, 2, 1, 3]);
        let reshape169_out1 = transpose343_out1.reshape([1, -1, 1024]);
        let linear219_out1 = self.linear219.forward(reshape169_out1);
        let add163_out1 = add160_out1.add(linear219_out1);
        let layernormalization141_out1 = {
            let dtype = add163_out1.clone().dtype();
            self.layernormalization141
                .forward(add163_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let squeeze48_out1 = self.squeeze48.val();
        let transpose344_out1 = squeeze48_out1.permute([1, 0]);
        let matmul290_out1 = layernormalization141_out1
            .matmul(transpose344_out1.unsqueeze::<3usize>());
        let split_tensors = matmul290_out1.split_with_sizes([1024, 1024].into(), 2);
        let [split24_out1, split24_out2] = split_tensors.try_into().unwrap();
        let sigmoid94_out1 = burn::tensor::activation::sigmoid(split24_out2);
        let mul184_out1 = split24_out1.mul(sigmoid94_out1);
        let transpose345_out1 = mul184_out1.permute([0, 2, 1]);
        let where72_out1 = transpose345_out1
            .mask_fill(unsqueeze33_out1, constant453_out1);
        let pad48_out1 = where72_out1
            .pad(
                [(0usize, 0usize), (0usize, 0usize), (0usize, 0usize)],
                burn::tensor::ops::PadMode::Constant(0f32),
            );
        let concat87_out1 = burn::tensor::Tensor::cat(
            [gather48_out1, pad48_out1].into(),
            2,
        );
        (
            concat87_out1,
            add163_out1,
            concat80_out1,
            concat83_out1,
            concat86_out1,
            slice90_out1,
            slice94_out1,
            slice98_out1,
        )
    }
}
#[derive(Module, Debug)]
pub struct Submodule25 {
    conv1d24: Conv1d,
    layernormalization142: LayerNorm,
    squeeze49: burn::module::Param<Tensor<2>>,
    layernormalization143: LayerNorm,
    linear220: Linear,
    linear221: Linear,
    layernormalization144: LayerNorm,
    #[module(skip)]
    device: Device,
}
impl Submodule25 {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let conv1d24 = Conv1dConfig::new(1024, 1024, 9)
            .with_stride(1)
            .with_padding(PaddingConfig1d::Valid)
            .with_dilation(1)
            .with_groups(1024)
            .with_bias(false)
            .init(device);
        let layernormalization142 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let squeeze49: burn::module::Param<Tensor<2>> = burn::module::Param::uninitialized(
            burn::module::ParamId::new(),
            move |device, _require_grad| Tensor::<
                2,
            >::zeros([1024, 1024], (device, burn::tensor::DType::F32)),
            device.clone(),
            false,
            [1024, 1024].into(),
        );
        let layernormalization143 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        let linear220 = LinearConfig::new(1024, 4096).with_bias(false).init(device);
        let linear221 = LinearConfig::new(4096, 1024).with_bias(false).init(device);
        let layernormalization144 = LayerNormConfig::new(1024)
            .with_epsilon(0.000009999999747378752f64)
            .with_bias(true)
            .init(device);
        Self {
            conv1d24,
            layernormalization142,
            squeeze49,
            layernormalization143,
            linear220,
            linear221,
            layernormalization144,
            device: device.clone(),
        }
    }
    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        concat87_out1: Tensor<3>,
        add163_out1: Tensor<3>,
        constant684_out1: Tensor<1>,
        concat15_out1: Tensor<3>,
        concat20_out1: Tensor<3>,
        concat23_out1: Tensor<3>,
        concat26_out1: Tensor<3>,
        concat29_out1: Tensor<3>,
        concat32_out1: Tensor<3>,
        concat35_out1: Tensor<3>,
        concat38_out1: Tensor<3>,
        concat41_out1: Tensor<3>,
        concat44_out1: Tensor<3>,
        concat47_out1: Tensor<3>,
        concat50_out1: Tensor<3>,
        concat53_out1: Tensor<3>,
        concat56_out1: Tensor<3>,
        concat59_out1: Tensor<3>,
        concat62_out1: Tensor<3>,
        concat65_out1: Tensor<3>,
        concat68_out1: Tensor<3>,
        concat71_out1: Tensor<3>,
        concat74_out1: Tensor<3>,
        concat77_out1: Tensor<3>,
        concat80_out1: Tensor<3>,
        concat83_out1: Tensor<3>,
        concat86_out1: Tensor<3>,
        slice10_out1: Tensor<3>,
        slice14_out1: Tensor<3>,
        slice18_out1: Tensor<3>,
        slice22_out1: Tensor<3>,
        slice26_out1: Tensor<3>,
        slice30_out1: Tensor<3>,
        slice34_out1: Tensor<3>,
        slice38_out1: Tensor<3>,
        slice42_out1: Tensor<3>,
        slice46_out1: Tensor<3>,
        slice50_out1: Tensor<3>,
        slice54_out1: Tensor<3>,
        slice58_out1: Tensor<3>,
        slice62_out1: Tensor<3>,
        slice66_out1: Tensor<3>,
        slice70_out1: Tensor<3>,
        slice74_out1: Tensor<3>,
        slice78_out1: Tensor<3>,
        slice82_out1: Tensor<3>,
        slice86_out1: Tensor<3>,
        slice90_out1: Tensor<3>,
        slice94_out1: Tensor<3>,
        slice98_out1: Tensor<3>,
        cache_last_channel_len: Tensor<1, Int>,
        max2_out1: i64,
        clip1_out1: Tensor<1, Int>,
    ) -> (Tensor<3>, Tensor<1, Int>, Tensor<4>, Tensor<4>, Tensor<1, Int>) {
        let slice102_out1 = concat87_out1.clone().slice(s![.., .., - 8..]);
        let conv1d24_out1 = self.conv1d24.forward(concat87_out1);
        let transpose346_out1 = conv1d24_out1.permute([0, 2, 1]);
        let layernormalization142_out1 = {
            let dtype = transpose346_out1.dtype();
            self.layernormalization142
                .forward(transpose346_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let sigmoid95_out1 = burn::tensor::activation::sigmoid(
            layernormalization142_out1.clone(),
        );
        let mul185_out1 = layernormalization142_out1.mul(sigmoid95_out1);
        let squeeze49_out1 = self.squeeze49.val();
        let transpose347_out1 = squeeze49_out1.permute([1, 0]);
        let matmul291_out1 = mul185_out1.matmul(transpose347_out1.unsqueeze::<3usize>());
        let add164_out1 = add163_out1.add(matmul291_out1);
        let layernormalization143_out1 = {
            let dtype = add164_out1.clone().dtype();
            self.layernormalization143
                .forward(add164_out1.clone().cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let linear220_out1 = self.linear220.forward(layernormalization143_out1);
        let sigmoid96_out1 = burn::tensor::activation::sigmoid(linear220_out1.clone());
        let mul186_out1 = linear220_out1.mul(sigmoid96_out1);
        let linear221_out1 = self.linear221.forward(mul186_out1);
        let mul187_out1 = linear221_out1
            .mul((constant684_out1).unsqueeze_dims(&[0isize, 1isize]));
        let add165_out1 = add164_out1.add(mul187_out1);
        let layernormalization144_out1 = {
            let dtype = add165_out1.dtype();
            self.layernormalization144
                .forward(add165_out1.cast(burn::tensor::DType::F32))
                .cast(dtype)
        };
        let transpose348_out1 = layernormalization144_out1.permute([0, 2, 1]);
        let unsqueeze34_out1: Tensor<4> = concat15_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze35_out1: Tensor<4> = concat20_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze36_out1: Tensor<4> = concat23_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze37_out1: Tensor<4> = concat26_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze38_out1: Tensor<4> = concat29_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze39_out1: Tensor<4> = concat32_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze40_out1: Tensor<4> = concat35_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze41_out1: Tensor<4> = concat38_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze42_out1: Tensor<4> = concat41_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze43_out1: Tensor<4> = concat44_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze44_out1: Tensor<4> = concat47_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze45_out1: Tensor<4> = concat50_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze46_out1: Tensor<4> = concat53_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze47_out1: Tensor<4> = concat56_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze48_out1: Tensor<4> = concat59_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze49_out1: Tensor<4> = concat62_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze50_out1: Tensor<4> = concat65_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze51_out1: Tensor<4> = concat68_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze52_out1: Tensor<4> = concat71_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze53_out1: Tensor<4> = concat74_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze54_out1: Tensor<4> = concat77_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze55_out1: Tensor<4> = concat80_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze56_out1: Tensor<4> = concat83_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze57_out1: Tensor<4> = concat86_out1.unsqueeze_dims::<4>(&[0]);
        let concat88_out1 = burn::tensor::Tensor::cat(
            [
                unsqueeze34_out1,
                unsqueeze35_out1,
                unsqueeze36_out1,
                unsqueeze37_out1,
                unsqueeze38_out1,
                unsqueeze39_out1,
                unsqueeze40_out1,
                unsqueeze41_out1,
                unsqueeze42_out1,
                unsqueeze43_out1,
                unsqueeze44_out1,
                unsqueeze45_out1,
                unsqueeze46_out1,
                unsqueeze47_out1,
                unsqueeze48_out1,
                unsqueeze49_out1,
                unsqueeze50_out1,
                unsqueeze51_out1,
                unsqueeze52_out1,
                unsqueeze53_out1,
                unsqueeze54_out1,
                unsqueeze55_out1,
                unsqueeze56_out1,
                unsqueeze57_out1,
            ]
                .into(),
            0,
        );
        let unsqueeze58_out1: Tensor<4> = slice10_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze59_out1: Tensor<4> = slice14_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze60_out1: Tensor<4> = slice18_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze61_out1: Tensor<4> = slice22_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze62_out1: Tensor<4> = slice26_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze63_out1: Tensor<4> = slice30_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze64_out1: Tensor<4> = slice34_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze65_out1: Tensor<4> = slice38_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze66_out1: Tensor<4> = slice42_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze67_out1: Tensor<4> = slice46_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze68_out1: Tensor<4> = slice50_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze69_out1: Tensor<4> = slice54_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze70_out1: Tensor<4> = slice58_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze71_out1: Tensor<4> = slice62_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze72_out1: Tensor<4> = slice66_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze73_out1: Tensor<4> = slice70_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze74_out1: Tensor<4> = slice74_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze75_out1: Tensor<4> = slice78_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze76_out1: Tensor<4> = slice82_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze77_out1: Tensor<4> = slice86_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze78_out1: Tensor<4> = slice90_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze79_out1: Tensor<4> = slice94_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze80_out1: Tensor<4> = slice98_out1.unsqueeze_dims::<4>(&[0]);
        let unsqueeze81_out1: Tensor<4> = slice102_out1.unsqueeze_dims::<4>(&[0]);
        let concat89_out1 = burn::tensor::Tensor::cat(
            [
                unsqueeze58_out1,
                unsqueeze59_out1,
                unsqueeze60_out1,
                unsqueeze61_out1,
                unsqueeze62_out1,
                unsqueeze63_out1,
                unsqueeze64_out1,
                unsqueeze65_out1,
                unsqueeze66_out1,
                unsqueeze67_out1,
                unsqueeze68_out1,
                unsqueeze69_out1,
                unsqueeze70_out1,
                unsqueeze71_out1,
                unsqueeze72_out1,
                unsqueeze73_out1,
                unsqueeze74_out1,
                unsqueeze75_out1,
                unsqueeze76_out1,
                unsqueeze77_out1,
                unsqueeze78_out1,
                unsqueeze79_out1,
                unsqueeze80_out1,
                unsqueeze81_out1,
            ]
                .into(),
            0,
        );
        let add166_out1 = cache_last_channel_len.add_scalar(max2_out1);
        let clip2_out1 = add166_out1.clamp_max(70f64);
        let slice103_out1 = concat88_out1.slice(s![.., .., - 70.., ..]);
        let slice104_out1 = transpose348_out1.slice(s![.., .., 0..14]);
        let clip3_out1 = clip1_out1.clamp_max(14f64);
        let transpose349_out1 = slice103_out1.permute([1, 0, 2, 3]);
        let transpose350_out1 = concat89_out1.permute([1, 0, 2, 3]);
        (slice104_out1, clip3_out1, transpose349_out1, transpose350_out1, clip2_out1)
    }
}

#[derive(Module, Debug)]
pub struct Model {
    submodule1: Submodule1,
    submodule2: Submodule2,
    submodule3: Submodule3,
    submodule4: Submodule4,
    submodule5: Submodule5,
    submodule6: Submodule6,
    submodule7: Submodule7,
    submodule8: Submodule8,
    submodule9: Submodule9,
    submodule10: Submodule10,
    submodule11: Submodule11,
    submodule12: Submodule12,
    submodule13: Submodule13,
    submodule14: Submodule14,
    submodule15: Submodule15,
    submodule16: Submodule16,
    submodule17: Submodule17,
    submodule18: Submodule18,
    submodule19: Submodule19,
    submodule20: Submodule20,
    submodule21: Submodule21,
    submodule22: Submodule22,
    submodule23: Submodule23,
    submodule24: Submodule24,
    submodule25: Submodule25,
    #[module(skip)]
    device: Device,
}





impl Model {
    #[allow(unused_variables)]
    pub fn new(device: &Device) -> Self {
        let submodule1 = Submodule1::new(device);
        let submodule2 = Submodule2::new(device);
        let submodule3 = Submodule3::new(device);
        let submodule4 = Submodule4::new(device);
        let submodule5 = Submodule5::new(device);
        let submodule6 = Submodule6::new(device);
        let submodule7 = Submodule7::new(device);
        let submodule8 = Submodule8::new(device);
        let submodule9 = Submodule9::new(device);
        let submodule10 = Submodule10::new(device);
        let submodule11 = Submodule11::new(device);
        let submodule12 = Submodule12::new(device);
        let submodule13 = Submodule13::new(device);
        let submodule14 = Submodule14::new(device);
        let submodule15 = Submodule15::new(device);
        let submodule16 = Submodule16::new(device);
        let submodule17 = Submodule17::new(device);
        let submodule18 = Submodule18::new(device);
        let submodule19 = Submodule19::new(device);
        let submodule20 = Submodule20::new(device);
        let submodule21 = Submodule21::new(device);
        let submodule22 = Submodule22::new(device);
        let submodule23 = Submodule23::new(device);
        let submodule24 = Submodule24::new(device);
        let submodule25 = Submodule25::new(device);
        Self {
            submodule1,
            submodule2,
            submodule3,
            submodule4,
            submodule5,
            submodule6,
            submodule7,
            submodule8,
            submodule9,
            submodule10,
            submodule11,
            submodule12,
            submodule13,
            submodule14,
            submodule15,
            submodule16,
            submodule17,
            submodule18,
            submodule19,
            submodule20,
            submodule21,
            submodule22,
            submodule23,
            submodule24,
            submodule25,
            device: device.clone(),
        }
    }

    #[allow(clippy::let_and_return, clippy::approx_constant)]
    pub fn forward(
        &self,
        processed_signal: Tensor<3>,
        processed_signal_length: Tensor<1, Int>,
        cache_last_channel: Tensor<4>,
        cache_last_time: Tensor<4>,
        cache_last_channel_len: Tensor<1, Int>,
        spk_targets: Tensor<2>,
        bg_spk_targets: Tensor<2>,
    ) -> (Tensor<3>, Tensor<1, Int>, Tensor<4>, Tensor<4>, Tensor<1, Int>) {
        let (
            sub7_out1,
            max2_out1,
            add6_out1,
            constant431_out1,
            constant427_out1,
            transpose1_out1,
            transpose2_out1,
            slice1_out1,
        ) = self
            .submodule1
            .forward(
                processed_signal,
                cache_last_channel,
                cache_last_time,
                processed_signal_length,
            );
        let (
            add21_out1,
            gather1_out1,
            transpose6_out1,
            unsqueeze19_out1,
            slice4_out1,
            slice3_out1,
            gather2_out1,
            clip1_out1,
        ) = self
            .submodule2
            .forward(
                sub7_out1,
                max2_out1.clone(),
                cache_last_channel_len.clone(),
                add6_out1.clone(),
                constant431_out1.clone(),
                constant427_out1,
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                spk_targets,
                slice1_out1,
                bg_spk_targets,
            );
        let (
            concat18_out1,
            add25_out1,
            constant684_out1,
            unsqueeze29_out1,
            concat16_out1,
            concat17_out1,
            constant687_out1,
            unsqueeze31_out1,
            constant450_out1,
            expand15_out1,
            constant453_out1,
            unsqueeze33_out1,
            concat15_out1,
        ) = self
            .submodule3
            .forward(
                add21_out1,
                gather1_out1,
                max2_out1.clone(),
                transpose6_out1.clone(),
                constant431_out1,
                add6_out1,
                unsqueeze19_out1.clone(),
                slice4_out1,
                slice3_out1,
                gather2_out1,
            );
        let (add31_out1, gather4_out1, concat20_out1, slice10_out1) = self
            .submodule4
            .forward(
                concat18_out1,
                add25_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
                constant453_out1.clone(),
            );
        let (add37_out1, gather6_out1, concat23_out1, slice14_out1) = self
            .submodule5
            .forward(
                add31_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather4_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add43_out1, gather8_out1, concat26_out1, slice18_out1) = self
            .submodule6
            .forward(
                add37_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather6_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add49_out1, gather10_out1, concat29_out1, slice22_out1) = self
            .submodule7
            .forward(
                add43_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather8_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add55_out1, gather12_out1, concat32_out1, slice26_out1) = self
            .submodule8
            .forward(
                add49_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather10_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add61_out1, gather14_out1, concat35_out1, slice30_out1) = self
            .submodule9
            .forward(
                add55_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather12_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add67_out1, gather16_out1, concat38_out1, slice34_out1) = self
            .submodule10
            .forward(
                add61_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather14_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add73_out1, gather18_out1, concat41_out1, slice38_out1) = self
            .submodule11
            .forward(
                add67_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather16_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add79_out1, gather20_out1, concat44_out1, slice42_out1) = self
            .submodule12
            .forward(
                add73_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather18_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add85_out1, gather22_out1, concat47_out1, slice46_out1) = self
            .submodule13
            .forward(
                add79_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather20_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add91_out1, gather24_out1, concat50_out1, slice50_out1) = self
            .submodule14
            .forward(
                add85_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather22_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add97_out1, gather26_out1, concat53_out1, slice54_out1) = self
            .submodule15
            .forward(
                add91_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather24_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add103_out1, gather28_out1, concat56_out1, slice58_out1) = self
            .submodule16
            .forward(
                add97_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather26_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add109_out1, gather30_out1, concat59_out1, slice62_out1) = self
            .submodule17
            .forward(
                add103_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather28_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add115_out1, gather32_out1, concat62_out1, slice66_out1) = self
            .submodule18
            .forward(
                add109_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather30_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add121_out1, gather34_out1, concat65_out1, slice70_out1) = self
            .submodule19
            .forward(
                add115_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather32_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add127_out1, gather36_out1, concat68_out1, slice74_out1) = self
            .submodule20
            .forward(
                add121_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather34_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add133_out1, gather38_out1, concat71_out1, slice78_out1) = self
            .submodule21
            .forward(
                add127_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather36_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add139_out1, gather40_out1, concat74_out1, slice82_out1) = self
            .submodule22
            .forward(
                add133_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather38_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (add145_out1, gather42_out1, concat77_out1, slice86_out1) = self
            .submodule23
            .forward(
                add139_out1,
                unsqueeze33_out1.clone(),
                constant453_out1.clone(),
                gather40_out1,
                constant684_out1.clone(),
                transpose1_out1.clone(),
                transpose2_out1.clone(),
                unsqueeze29_out1.clone(),
                transpose6_out1.clone(),
                concat16_out1.clone(),
                concat17_out1.clone(),
                unsqueeze19_out1.clone(),
                constant687_out1.clone(),
                unsqueeze31_out1.clone(),
                constant450_out1.clone(),
                expand15_out1.clone(),
            );
        let (
            concat87_out1,
            add163_out1,
            concat80_out1,
            concat83_out1,
            concat86_out1,
            slice90_out1,
            slice94_out1,
            slice98_out1,
        ) = self
            .submodule24
            .forward(
                add145_out1,
                unsqueeze33_out1,
                constant453_out1,
                gather42_out1,
                constant684_out1.clone(),
                transpose1_out1,
                transpose2_out1,
                unsqueeze29_out1,
                transpose6_out1,
                concat16_out1,
                concat17_out1,
                unsqueeze19_out1,
                constant687_out1,
                unsqueeze31_out1,
                constant450_out1,
                expand15_out1,
            );
        let (
            slice104_out1,
            clip3_out1,
            transpose349_out1,
            transpose350_out1,
            clip2_out1,
        ) = self
            .submodule25
            .forward(
                concat87_out1,
                add163_out1,
                constant684_out1,
                concat15_out1,
                concat20_out1,
                concat23_out1,
                concat26_out1,
                concat29_out1,
                concat32_out1,
                concat35_out1,
                concat38_out1,
                concat41_out1,
                concat44_out1,
                concat47_out1,
                concat50_out1,
                concat53_out1,
                concat56_out1,
                concat59_out1,
                concat62_out1,
                concat65_out1,
                concat68_out1,
                concat71_out1,
                concat74_out1,
                concat77_out1,
                concat80_out1,
                concat83_out1,
                concat86_out1,
                slice10_out1,
                slice14_out1,
                slice18_out1,
                slice22_out1,
                slice26_out1,
                slice30_out1,
                slice34_out1,
                slice38_out1,
                slice42_out1,
                slice46_out1,
                slice50_out1,
                slice54_out1,
                slice58_out1,
                slice62_out1,
                slice66_out1,
                slice70_out1,
                slice74_out1,
                slice78_out1,
                slice82_out1,
                slice86_out1,
                slice90_out1,
                slice94_out1,
                slice98_out1,
                cache_last_channel_len,
                max2_out1,
                clip1_out1,
            );
        (slice104_out1, clip3_out1, transpose349_out1, transpose350_out1, clip2_out1)
    }
}