burn_ir/

operation.rs

use burn_backend::ops::AttentionModuleOptions;
use burn_backend::tensor::IndexingUpdateOp;
use core::hash::Hash;
use serde::{Deserialize, Serialize};

use alloc::borrow::ToOwned;
use alloc::boxed::Box;
use alloc::{string::String, vec::Vec};

use burn_backend::{
    DType, Distribution, Slice,
    ops::{
        ConvOptions, ConvTransposeOptions, DeformConvOptions, GridSampleOptions,
        GridSamplePaddingMode, InterpolateMode, InterpolateOptions,
    },
    quantization::QuantScheme,
};

use crate::{ScalarIr, TensorId, TensorIr, TensorStatus};

/// Visitor for mutating the components of an [`OperationIr`] in place.
pub trait IrVisitorMut {
    /// Visit a [`TensorIr`] mutably.
    fn visit_tensor_mut(&mut self, _tensor: &mut TensorIr) {}
    /// Visit a [`ScalarIr`] mutably.
    fn visit_scalar_mut(&mut self, _scalar: &mut ScalarIr) {}
    /// Visit a slice range mutably.
    fn visit_range_mut(&mut self, _range: &mut Slice) {}
}

/// Custom operation in fusion stream, declaring its inputs, outputs and scalar arguments.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct CustomOpIr {
    /// Unique identifier of the operation.
    pub id: String,
    /// Input tensors used in the custom operation.
    pub inputs: Vec<TensorIr>,
    /// Output tensors used in the custom operation.
    pub outputs: Vec<TensorIr>,
    /// Non-tensor scalar arguments, in declaration order.
    ///
    /// Carried through fusion's relativization like any other scalar (see the
    /// `RelativeOps for CustomOpIr` impl), so a cached graph replays with fresh scalar values. The
    /// remote backend relies on these to ship a custom op's scalar arguments to the server, where
    /// the registered handler reads them back with [`ScalarIr::elem`].
    pub scalars: Vec<ScalarIr>,
}

impl CustomOpIr {
    /// Create a new custom operation intermediate representation (without scalar arguments).
    pub fn new(id: &'static str, inputs: &[TensorIr], outputs: &[TensorIr]) -> Self {
        Self {
            id: id.to_owned(),
            inputs: inputs.to_vec(),
            outputs: outputs.to_vec(),
            scalars: Vec::new(),
        }
    }

    /// Create a new custom operation intermediate representation with scalar arguments.
    pub fn with_scalars(
        id: &'static str,
        inputs: &[TensorIr],
        outputs: &[TensorIr],
        scalars: Vec<ScalarIr>,
    ) -> Self {
        Self {
            id: id.to_owned(),
            inputs: inputs.to_vec(),
            outputs: outputs.to_vec(),
            scalars,
        }
    }

    /// Cast the intermediate representation, and get the in and output tensors.
    pub fn as_fixed<const N_IN: usize, const N_OUT: usize>(
        &self,
    ) -> (&[TensorIr; N_IN], &[TensorIr; N_OUT]) {
        (
            self.inputs.as_slice().try_into().expect(
                "Wrong number of inputs expected (expected {D}, is {}), check your implementation",
            ),
            self.outputs.as_slice().try_into().expect(
                "Wrong number of outputs expected (expected {D}, is {}), check your implementation",
            ),
        )
    }

    fn inputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        Box::new(self.inputs.iter())
    }

    fn outputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        Box::new(self.outputs.iter())
    }

    fn visit_mut(&mut self, v: &mut impl IrVisitorMut) {
        for t in self.inputs.iter_mut() {
            v.visit_tensor_mut(t);
        }
        for t in self.outputs.iter_mut() {
            v.visit_tensor_mut(t);
        }
        for s in self.scalars.iter_mut() {
            v.visit_scalar_mut(s);
        }
    }
}

/// Describe all tensor operations possible.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(clippy::large_enum_variant)]
pub enum OperationIr {
    /// Basic operation on a float tensor.
    BaseFloat(BaseOperationIr),
    /// Basic operation on an int tensor.
    BaseInt(BaseOperationIr),
    /// Basic operation on a bool tensor.
    BaseBool(BaseOperationIr),
    /// Numeric operation on a float tensor.
    NumericFloat(DType, NumericOperationIr),
    /// Numeric operation on an int tensor.
    NumericInt(DType, NumericOperationIr),
    /// Operation specific to a bool tensor.
    Bool(BoolOperationIr),
    /// Operation specific to an int tensor.
    Int(IntOperationIr),
    /// Operation specific to a float tensor.
    Float(DType, FloatOperationIr),
    /// Module operation.
    Module(ModuleOperationIr),
    /// Initialize operation.
    Init(InitOperationIr),
    /// A custom operation.
    Custom(CustomOpIr),
    /// A tensor is dropped.
    Drop(TensorIr),
    /// Operation specific to a distributed tensor.
    Distributed(DistributedOperationIr),
    /// Activation function operation (relu, gelu, sigmoid, softmax, …).
    Activation(ActivationOperationIr),
}

/// Operation intermediate representation specific to a float tensor.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub enum FloatOperationIr {
    /// Operation corresponding to [exp](burn_backend::ops::FloatTensorOps::float_exp).
    Exp(UnaryOpIr),
    /// Operation corresponding to [log](burn_backend::ops::FloatTensorOps::float_log).
    Log(UnaryOpIr),
    /// Operation corresponding to [log1p](burn_backend::ops::FloatTensorOps::float_log1p).
    Log1p(UnaryOpIr),
    /// Operation corresponding to [erf](burn_backend::ops::FloatTensorOps::float_erf).
    Erf(UnaryOpIr),
    /// Operation corresponding to [powf_scalar](burn_backend::ops::FloatTensorOps::float_powf_scalar).
    PowfScalar(ScalarOpIr),
    /// Operation corresponding to [sqrt](burn_backend::ops::FloatTensorOps::float_sqrt).
    Sqrt(UnaryOpIr),
    /// Operation corresponding to [cos](burn_backend::ops::FloatTensorOps::float_cos).
    Cos(UnaryOpIr),
    /// Operation corresponding to [cosh](burn_backend::ops::FloatTensorOps::float_cosh).
    Cosh(UnaryOpIr),
    /// Operation corresponding to [sin](burn_backend::ops::FloatTensorOps::float_sin).
    Sin(UnaryOpIr),
    /// Operation corresponding to [sin](burn_backend::ops::FloatTensorOps::float_sinh).
    Sinh(UnaryOpIr),
    /// Operation corresponding to [tan](burn_backend::ops::FloatTensorOps::float_tan).
    Tan(UnaryOpIr),
    /// Operation corresponding to [tanh](burn_backend::ops::FloatTensorOps::float_tanh).
    Tanh(UnaryOpIr),
    /// Operation corresponding to [acos](burn_backend::ops::FloatTensorOps::float_acos).
    ArcCos(UnaryOpIr),
    /// Operation corresponding to [acosh](burn_backend::ops::FloatTensorOps::float_acosh).
    ArcCosh(UnaryOpIr),
    /// Operation corresponding to [asin](burn_backend::ops::FloatTensorOps::float_asin).
    ArcSin(UnaryOpIr),
    /// Operation corresponding to [asinh](burn_backend::ops::FloatTensorOps::float_asinh).
    ArcSinh(UnaryOpIr),
    /// Operation corresponding to [atan](burn_backend::ops::FloatTensorOps::float_atan).
    ArcTan(UnaryOpIr),
    /// Operation corresponding to [atanh](burn_backend::ops::FloatTensorOps::float_atanh).
    ArcTanh(UnaryOpIr),
    /// Operation corresponding to [atan2](burn_backend::ops::FloatTensorOps::float_atan2).
    ArcTan2(BinaryOpIr),
    /// Operation corresponding to [round](burn_backend::ops::FloatTensorOps::float_round).
    Round(UnaryOpIr),
    /// Operation corresponding to [floor](burn_backend::ops::FloatTensorOps::float_floor).
    Floor(UnaryOpIr),
    /// Operation corresponding to [ceil](burn_backend::ops::FloatTensorOps::float_ceil).
    Ceil(UnaryOpIr),
    /// Operation corresponding to [trunc](burn_backend::ops::FloatTensorOps::float_trunc).
    Trunc(UnaryOpIr),
    /// Operation corresponding to [into_int](burn_backend::ops::FloatTensorOps::float_into_int).
    IntoInt(CastOpIr),
    /// Operation corresponding to [matmul](burn_backend::ops::FloatTensorOps::float_matmul).
    Matmul(MatmulOpIr),
    /// Operation corresponding to [cross](burn_backend::ops::FloatTensorOps::float_cross).
    Cross(CrossOpIr),
    /// Operation corresponding to [random](burn_backend::ops::FloatTensorOps::float_random).
    Random(RandomOpIr),
    /// Operation corresponding to [recip](burn_backend::ops::FloatTensorOps::float_recip).
    Recip(UnaryOpIr),
    /// Operation corresponding to [is_nan](burn_backend::ops::FloatTensorOps::float_is_nan).
    IsNan(UnaryOpIr),
    /// Operation corresponding to [is_nan](burn_backend::ops::FloatTensorOps::float_is_inf).
    IsInf(UnaryOpIr),
    /// Operation corresponding to [quantize](burn_backend::ops::QTensorOps::quantize).
    Quantize(QuantizeOpIr),
    /// Operation corresponding to [dequantize](burn_backend::ops::QTensorOps::dequantize).
    Dequantize(DequantizeOpIr),
    /// Operation corresponding to [grid_sample_2d](burn_backend::ops::FloatTensorOps::float_grid_sample_2d).
    GridSample2d(GridSample2dOpIr),
    /// Operation corresponding to [powf](burn_backend::ops::FloatTensorOps::float_powi).
    Powf(BinaryOpIr),
    /// Operation corresponding to [hypot](burn_backend::ops::FloatTensorOps::float_hypot).
    Hypot(BinaryOpIr),
}

/// Operation intermediate representation specific to module.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub enum ModuleOperationIr {
    /// Batch normalization with explicitly supplied statistics, corresponding
    /// to [batch_norm](burn_backend::ops::ModuleOps::batch_norm).
    BatchNorm(BatchNormOpIr),
    /// Operation corresponding to [embedding](burn_backend::ops::ModuleOps::embedding).
    Embedding(EmbeddingOpIr),
    /// Operation corresponding to [embedding_backward](burn_backend::ops::ModuleOps::embedding_backward).
    EmbeddingBackward(EmbeddingBackwardOpIr),
    /// Operation corresponding to [linear](burn_backend::ops::ModuleOps::linear).
    Linear(LinearOpIr),
    /// Operation corresponding to [linear_x_backward](burn_backend::ops::ModuleOps::linear_x_backward).
    LinearXBackward(LinearXBackwardOpIr),
    /// Operation corresponding to [linear_weight_backward](burn_backend::ops::ModuleOps::linear_weight_backward).
    LinearWeightBackward(LinearWeightBackwardOpIr),
    /// Operation corresponding to [linear_bias_backward](burn_backend::ops::ModuleOps::linear_bias_backward).
    LinearBiasBackward(LinearBiasBackwardOpIr),
    /// Operation corresponding to [conv1d](burn_backend::ops::ModuleOps::conv1d).
    Conv1d(Conv1dOpIr),
    /// Operation corresponding to [conv1d_x_backward](burn_backend::ops::ModuleOps::conv1d_x_backward).
    Conv1dXBackward(Conv1dXBackwardOpIr),
    /// Operation corresponding to [conv1d_weight_backward](burn_backend::ops::ModuleOps::conv1d_weight_backward).
    Conv1dWeightBackward(Conv1dWeightBackwardOpIr),
    /// Operation corresponding to [conv1d_bias_backward](burn_backend::ops::ModuleOps::conv1d_bias_backward).
    Conv1dBiasBackward(Conv1dBiasBackwardOpIr),
    /// Operation corresponding to [conv2d](burn_backend::ops::ModuleOps::conv2d).
    Conv2d(Conv2dOpIr),
    /// Operation corresponding to [conv2d_x_backward](burn_backend::ops::ModuleOps::conv2d_x_backward).
    Conv2dXBackward(Conv2dXBackwardOpIr),
    /// Operation corresponding to [conv2d_weight_backward](burn_backend::ops::ModuleOps::conv2d_weight_backward).
    Conv2dWeightBackward(Conv2dWeightBackwardOpIr),
    /// Operation corresponding to [conv2d_bias_backward](burn_backend::ops::ModuleOps::conv2d_bias_backward).
    Conv2dBiasBackward(Conv2dBiasBackwardOpIr),
    /// Operation corresponding to [conv3d](burn_backend::ops::ModuleOps::conv3d).
    Conv3d(Conv3dOpIr),
    /// Operation corresponding to [conv3d_x_backward](burn_backend::ops::ModuleOps::conv3d_x_backward).
    Conv3dXBackward(Conv3dXBackwardOpIr),
    /// Operation corresponding to [conv3d_weight_backward](burn_backend::ops::ModuleOps::conv3d_weight_backward).
    Conv3dWeightBackward(Conv3dWeightBackwardOpIr),
    /// Operation corresponding to [conv3d_bias_backward](burn_backend::ops::ModuleOps::conv3d_bias_backward).
    Conv3dBiasBackward(Conv3dBiasBackwardOpIr),
    /// Operation corresponding to [deform_conv2d](burn_backend::ops::ModuleOps::deform_conv2d)
    DeformableConv2d(Box<DeformConv2dOpIr>),
    /// Operation corresponding to [deform_conv2d_backward](burn_backend::ops::ModuleOps::deform_conv2d_backward)
    DeformableConv2dBackward(Box<DeformConv2dBackwardOpIr>),
    /// Operation corresponding to [conv transpose 1d](burn_backend::ops::ModuleOps::conv_transpose1d).
    ConvTranspose1d(ConvTranspose1dOpIr),
    /// Operation corresponding to [conv transpose 2d](burn_backend::ops::ModuleOps::conv_transpose2d).
    ConvTranspose2d(ConvTranspose2dOpIr),
    /// Operation corresponding to [conv transpose 3d](burn_backend::ops::ModuleOps::conv_transpose3d).
    ConvTranspose3d(ConvTranspose3dOpIr),
    /// Operation corresponding to [avg pool 1d](burn_backend::ops::ModuleOps::avg_pool1d).
    AvgPool1d(AvgPool1dOpIr),
    /// Operation corresponding to [avg pool 2d](burn_backend::ops::ModuleOps::avg_pool2d).
    AvgPool2d(AvgPool2dOpIr),
    /// Operation corresponding to
    /// [avg pool 1d backward](burn_backend::ops::ModuleOps::avg_pool1d_backward).
    AvgPool1dBackward(AvgPool1dBackwardOpIr),
    /// Operation corresponding to
    /// [avg pool 2d backward](burn_backend::ops::ModuleOps::avg_pool2d_backward).
    AvgPool2dBackward(AvgPool2dBackwardOpIr),
    /// Operation corresponding to
    /// [adaptive avg pool 1d](burn_backend::ops::ModuleOps::adaptive_avg_pool1d).
    AdaptiveAvgPool1d(AdaptiveAvgPool1dOpIr),
    /// Operation corresponding to
    /// [adaptive avg pool 2d](burn_backend::ops::ModuleOps::adaptive_avg_pool2d).
    AdaptiveAvgPool2d(AdaptiveAvgPool2dOpIr),
    /// Operation corresponding to
    /// [adaptive avg pool 1d backward](burn_backend::ops::ModuleOps::adaptive_avg_pool1d_backward).
    AdaptiveAvgPool1dBackward(AdaptiveAvgPool1dBackwardOpIr),
    /// Operation corresponding to
    /// [adaptive avg pool 2d backward](burn_backend::ops::ModuleOps::adaptive_avg_pool2d_backward).
    AdaptiveAvgPool2dBackward(AdaptiveAvgPool2dBackwardOpIr),
    /// Operation corresponding to
    /// [adaptive avg pool 3d](burn_backend::ops::ModuleOps::adaptive_avg_pool3d).
    AdaptiveAvgPool3d(AdaptiveAvgPool3dOpIr),
    /// Operation corresponding to
    /// [adaptive avg pool 3d backward](burn_backend::ops::ModuleOps::adaptive_avg_pool3d_backward).
    AdaptiveAvgPool3dBackward(AdaptiveAvgPool3dBackwardOpIr),
    /// Operation corresponding to
    /// [max pool 1d](burn_backend::ops::ModuleOps::max_pool1d).
    MaxPool1d(MaxPool1dOpIr),
    /// Operation corresponding to
    /// [max pool 1d with indices](burn_backend::ops::ModuleOps::max_pool1d_with_indices).
    MaxPool1dWithIndices(MaxPool1dWithIndicesOpIr),
    /// Operation corresponding to
    /// [max pool 1d with indices backward](burn_backend::ops::ModuleOps::max_pool1d_with_indices_backward).
    MaxPool1dWithIndicesBackward(MaxPool1dWithIndicesBackwardOpIr),
    /// Operation corresponding to
    /// [max pool 2d](burn_backend::ops::ModuleOps::max_pool1d).
    MaxPool2d(MaxPool2dOpIr),
    /// Operation corresponding to
    /// [max pool 2d with indices](burn_backend::ops::ModuleOps::max_pool2d_with_indices).
    MaxPool2dWithIndices(MaxPool2dWithIndicesOpIr),
    /// Operation corresponding to
    /// [max pool 2d with indices backward](burn_backend::ops::ModuleOps::max_pool2d_with_indices_backward).
    MaxPool2dWithIndicesBackward(MaxPool2dWithIndicesBackwardOpIr),
    /// Operation corresponding to [interpolate](burn_backend::ops::ModuleOps::interpolate).
    Interpolate(InterpolateOpIr),
    /// Operation corresponding to [interpolate backward](burn_backend::ops::ModuleOps::interpolate_backward).
    InterpolateBackward(InterpolateBackwardOpIr),
    /// Operation corresponding to [rfft](burn_backend::ops::ModuleOps::rfft)
    Rfft(RfftOpIr),
    /// Operation corresponding to [irfft](burn_backend::ops::ModuleOps::irfft)
    IRfft(IRfftOpIr),
    /// Operation corresponding to [attention](burn_backend::ops::ModuleOps::attention).
    Attention(AttentionOpIr),
    /// Operation corresponding to [ctc_loss](burn_backend::ops::ModuleOps::ctc_loss).
    CtcLoss(CtcLossOpIr),
    /// Operation corresponding to
    /// [ctc_loss_backward](burn_backend::ops::ModuleOps::ctc_loss_backward).
    CtcLossBackward(CtcLossBackwardOpIr),
    /// Operation corresponding to [layer_norm](burn_backend::ops::ModuleOps::layer_norm).
    LayerNorm(LayerNormOpIr),
    /// Operation corresponding to [unfold4d](burn_backend::ops::ModuleOps::unfold4d).
    Unfold4d(Unfold4dOpIr),
    /// Operation corresponding to
    /// [conv_transpose1d_weight_backward](burn_backend::ops::ModuleOps::conv_transpose1d_weight_backward).
    ConvTranspose1dWeightBackward(ConvTranspose1dWeightBackwardOpIr),
    /// Operation corresponding to
    /// [conv_transpose1d_bias_backward](burn_backend::ops::ModuleOps::conv_transpose1d_bias_backward).
    ConvTranspose1dBiasBackward(ConvTranspose1dBiasBackwardOpIr),
    /// Operation corresponding to
    /// [conv_transpose2d_weight_backward](burn_backend::ops::ModuleOps::conv_transpose2d_weight_backward).
    ConvTranspose2dWeightBackward(ConvTranspose2dWeightBackwardOpIr),
    /// Operation corresponding to
    /// [conv_transpose2d_bias_backward](burn_backend::ops::ModuleOps::conv_transpose2d_bias_backward).
    ConvTranspose2dBiasBackward(ConvTranspose2dBiasBackwardOpIr),
    /// Operation corresponding to
    /// [conv_transpose3d_weight_backward](burn_backend::ops::ModuleOps::conv_transpose3d_weight_backward).
    ConvTranspose3dWeightBackward(ConvTranspose3dWeightBackwardOpIr),
    /// Operation corresponding to
    /// [conv_transpose3d_bias_backward](burn_backend::ops::ModuleOps::conv_transpose3d_bias_backward).
    ConvTranspose3dBiasBackward(ConvTranspose3dBiasBackwardOpIr),
}

/// Basic operations that can be done on any tensor type.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub enum BaseOperationIr {
    /// Operation corresponding to:
    ///
    /// Float => [reshape](burn_backend::ops::FloatTensorOps::float_reshape).
    /// Int => [reshape](burn_backend::ops::IntTensorOps::int_reshape).
    /// Bool => [reshape](burn_backend::ops::BoolTensorOps::bool_reshape).
    Reshape(ShapeOpIr),

    /// Operation corresponding to:
    ///
    /// Float => [swap_dims](burn_backend::ops::FloatTensorOps::float_swap_dims).
    /// Int => [swap_dims](burn_backend::ops::IntTensorOps::int_swap_dims).
    /// Bool => [swap_dims](burn_backend::ops::BoolTensorOps::bool_swap_dims).
    SwapDims(SwapDimsOpIr),

    /// Operation corresponding to:
    ///
    /// Float => [permute](burn_backend::ops::FloatTensorOps::float_permute).
    /// Int => [permute](burn_backend::ops::IntTensorOps::int_permute).
    /// Bool => [permute](burn_backend::ops::BoolTensorOps::bool_permute).
    Permute(PermuteOpIr),

    /// Operation corresponding to:
    /// Float => [flip](burn_backend::ops::FloatTensorOps::float_flip).
    /// Int => [flip](burn_backend::ops::IntTensorOps::int_flip).
    /// Bool => [flip](burn_backend::ops::BoolTensorOps::bool_flip).
    Flip(FlipOpIr),

    /// Operation corresponding to:
    ///
    /// Float => [expand](burn_backend::ops::FloatTensorOps::float_expand).
    /// Int => [expand](burn_backend::ops::IntTensorOps::int_expand).
    /// Bool => [expand](burn_backend::ops::BoolTensorOps::bool_expand).
    Expand(ShapeOpIr),

    /// Unfold windows along an axis.
    ///
    Unfold(UnfoldOpIr),

    /// Operation corresponding to:
    ///
    /// Float => [slice](burn_backend::ops::FloatTensorOps::float_slice).
    /// Int => [slice](burn_backend::ops::IntTensorOps::int_slice).
    /// Bool => [slice](burn_backend::ops::BoolTensorOps::bool_slice).
    Slice(SliceOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [slice assign](burn_backend::ops::FloatTensorOps::float_slice_assign).
    /// Int => [slice assign](burn_backend::ops::IntTensorOps::int_slice_assign).
    /// Bool => [slice assign](burn_backend::ops::BoolTensorOps::bool_slice_assign).
    SliceAssign(SliceAssignOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [select](burn_backend::ops::FloatTensorOps::float_select).
    /// Int => [select](burn_backend::ops::IntTensorOps::int_select).
    /// Bool => [select](burn_backend::ops::BoolTensorOps::bool_select).
    Select(SelectOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [select assign](burn_backend::ops::FloatTensorOps::float_select_assign).
    /// Int => [select assign](burn_backend::ops::IntTensorOps::int_select_assign).
    /// Bool => [select assign](burn_backend::ops::BoolTensorOps::bool_select_or).
    SelectAssign(SelectAssignOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [mask where](burn_backend::ops::FloatTensorOps::float_mask_where).
    /// Int => [mask where](burn_backend::ops::IntTensorOps::int_mask_where).
    /// Bool => [mask where](burn_backend::ops::BoolTensorOps::bool_mask_where).
    MaskWhere(MaskWhereOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [mask fill](burn_backend::ops::FloatTensorOps::float_mask_fill).
    /// Int => [mask fill](burn_backend::ops::IntTensorOps::int_mask_fill).
    /// Bool => [mask fill](burn_backend::ops::BoolTensorOps::bool_mask_fill).
    MaskFill(MaskFillOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [gather](burn_backend::ops::FloatTensorOps::float_gather).
    /// Int => [gather](burn_backend::ops::IntTensorOps::int_gather).
    /// Bool => [gather](burn_backend::ops::BoolTensorOps::bool_gather).
    Gather(GatherOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [scatter](burn_backend::ops::FloatTensorOps::float_scatter_add).
    /// Int => [scatter](burn_backend::ops::IntTensorOps::int_scatter_add).
    /// Bool => [scatter](burn_backend::ops::BoolTensorOps::bool_scatter_or).
    Scatter(ScatterOpIr),
    /// Multi-dimensional scatter operation.
    ScatterNd(ScatterNdOpIr),
    /// Multi-dimensional gather operation.
    GatherNd(GatherNdOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [equal](burn_backend::ops::FloatTensorOps::float_equal).
    /// Int => [equal](burn_backend::ops::IntTensorOps::int_equal).
    /// Bool => [equal](burn_backend::ops::BoolTensorOps::bool_equal).
    Equal(BinaryOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [equal elem](burn_backend::ops::FloatTensorOps::float_equal_elem).
    /// Int => [equal elem](burn_backend::ops::IntTensorOps::int_equal_elem).
    /// Bool => [equal elem](burn_backend::ops::BoolTensorOps::bool_equal_elem).
    EqualElem(ScalarOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [repeat dim](burn_backend::ops::FloatTensorOps::float_repeat_dim).
    /// Int => [repeat dim](burn_backend::ops::IntTensorOps::int_repeat_dim).
    /// Bool => [repeat dim](burn_backend::ops::BoolTensorOps::bool_repeat_dim).
    RepeatDim(RepeatDimOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [cat](burn_backend::ops::FloatTensorOps::float_cat).
    /// Int => [cat](burn_backend::ops::IntTensorOps::int_cat).
    /// Bool => [cat](burn_backend::ops::BoolTensorOps::bool_cat).
    Cat(CatOpIr),
    /// Cast operation, no direct operation and should be supported by fusion backend.
    Cast(CastOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [empty](burn_backend::ops::FloatTensorOps::float_empty).
    /// Int => [empty](burn_backend::ops::IntTensorOps::int_empty).
    /// Bool => [empty](burn_backend::ops::BoolTensorOps::bool_empty).
    Empty(CreationOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [ones](burn_backend::ops::FloatTensorOps::float_ones).
    /// Int => [ones](burn_backend::ops::IntTensorOps::int_ones).
    /// Bool => [ones](burn_backend::ops::BoolTensorOps::bool_ones).
    Ones(CreationOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [zeros](burn_backend::ops::FloatTensorOps::float_zeros).
    /// Int => [zeros](burn_backend::ops::IntTensorOps::int_zeros).
    /// Bool => [zeros](burn_backend::ops::BoolTensorOps::bool_zeros).
    Zeros(CreationOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [not_equal](burn_backend::ops::FloatTensorOps::float_not_equal).
    /// Int => [not_equal](burn_backend::ops::IntTensorOps::int_not_equal).
    /// Bool => [not_equal](burn_backend::ops::BoolTensorOps::bool_not_equal).
    NotEqual(BinaryOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [not_equal_elem](burn_backend::ops::FloatTensorOps::float_not_equal_elem).
    /// Int => [not_equal_elem](burn_backend::ops::IntTensorOps::int_not_equal_elem).
    /// Bool => [not_equal_elem](burn_backend::ops::BoolTensorOps::bool_not_equal_elem).
    NotEqualElem(ScalarOpIr),
    /// Reduce-`all` over the input tensor.
    ///
    /// Float/Int input is treated as a non-zero check; bool input is the value itself.
    /// Output is a single-element bool tensor.
    All(ReduceOpIr),
    /// Reduce-`any` over the input tensor.
    Any(ReduceOpIr),
    /// Reduce-`all` along a dim.
    AllDim(ReduceDimOpIr),
    /// Reduce-`any` along a dim.
    AnyDim(ReduceDimOpIr),
}

/// Numeric operations on int and float tensors.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub enum NumericOperationIr {
    /// Operation corresponding to:
    ///
    /// Float => [add](burn_backend::ops::FloatTensorOps::float_add).
    /// Int => [add](burn_backend::ops::IntTensorOps::int_add).
    Add(BinaryOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [add scalar](burn_backend::ops::FloatTensorOps::float_add_scalar).
    /// Int => [add scalar](burn_backend::ops::IntTensorOps::int_add_scalar).
    AddScalar(ScalarOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [sub](burn_backend::ops::FloatTensorOps::float_sub).
    /// Int => [sub](burn_backend::ops::IntTensorOps::int_sub).
    Sub(BinaryOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [sub scalar](burn_backend::ops::FloatTensorOps::float_sub_scalar).
    /// Int => [sub scalar](burn_backend::ops::IntTensorOps::int_sub_scalar).
    SubScalar(ScalarOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [div](burn_backend::ops::FloatTensorOps::float_div).
    /// Int => [div](burn_backend::ops::IntTensorOps::int_div).
    Div(BinaryOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [div scalar](burn_backend::ops::FloatTensorOps::float_div_scalar).
    /// Int => [div scalar](burn_backend::ops::IntTensorOps::int_div_scalar).
    DivScalar(ScalarOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [rem](burn_backend::ops::FloatTensorOps::float_remainder).
    /// Int => [rem](burn_backend::ops::IntTensorOps::int_remainder).
    Rem(BinaryOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [rem scalar](burn_backend::ops::FloatTensorOps::float_remainder_scalar).
    /// Int => [rem scalar](burn_backend::ops::IntTensorOps::int_remainder_scalar).
    RemScalar(ScalarOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [mul](burn_backend::ops::FloatTensorOps::float_mul).
    /// Int => [mul](burn_backend::ops::IntTensorOps::int_mul).
    Mul(BinaryOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [mul scalar](burn_backend::ops::FloatTensorOps::float_mul_scalar).
    /// Int => [mul scalar](burn_backend::ops::IntTensorOps::int_mul_scalar).
    MulScalar(ScalarOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [abs](burn_backend::ops::FloatTensorOps::float_abs).
    /// Int => [abs](burn_backend::ops::IntTensorOps::int_abs).
    Abs(UnaryOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [full](burn_backend::ops::FloatTensorOps::float_full).
    /// Int => [full](burn_backend::ops::IntTensorOps::int_full).
    Full(FullOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [mean dim](burn_backend::ops::FloatTensorOps::float_mean_dim).
    /// Int => [mean dim](burn_backend::ops::IntTensorOps::int_mean_dim).
    MeanDim(ReduceDimOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [mean](burn_backend::ops::FloatTensorOps::float_mean).
    /// Int => [mean](burn_backend::ops::IntTensorOps::int_mean).
    Mean(ReduceOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [sum](burn_backend::ops::FloatTensorOps::float_sum).
    /// Int => [sum](burn_backend::ops::IntTensorOps::int_sum).
    Sum(ReduceOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [sum dim](burn_backend::ops::FloatTensorOps::float_sum_dim).
    /// Int => [sum dim](burn_backend::ops::IntTensorOps::int_sum_dim).
    SumDim(ReduceDimOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [prod](burn_backend::ops::FloatTensorOps::float_prod).
    /// Int => [prod](burn_backend::ops::IntTensorOps::int_prod).
    Prod(ReduceOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [prod dim](burn_backend::ops::FloatTensorOps::float_prod_dim).
    /// Int => [prod dim](burn_backend::ops::IntTensorOps::int_prod_dim).
    ProdDim(ReduceDimOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [greater](burn_backend::ops::FloatTensorOps::float_greater).
    /// Int => [greater](burn_backend::ops::IntTensorOps::int_greater).
    Greater(BinaryOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [greater elem](burn_backend::ops::FloatTensorOps::float_greater_elem).
    /// Int => [greater elem](burn_backend::ops::IntTensorOps::int_greater_elem).
    GreaterElem(ScalarOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [greater equal](burn_backend::ops::FloatTensorOps::float_greater_elem).
    /// Int => [greater elem](burn_backend::ops::IntTensorOps::int_greater_elem).
    GreaterEqual(BinaryOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [greater equal elem](burn_backend::ops::FloatTensorOps::float_greater_equal_elem).
    /// Int => [greater equal elem](burn_backend::ops::IntTensorOps::int_greater_equal_elem).
    GreaterEqualElem(ScalarOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [lower](burn_backend::ops::FloatTensorOps::float_lower).
    /// Int => [lower](burn_backend::ops::IntTensorOps::int_lower).
    Lower(BinaryOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [lower elem](burn_backend::ops::FloatTensorOps::float_lower_elem).
    /// Int => [lower elem](burn_backend::ops::IntTensorOps::int_lower_elem).
    LowerElem(ScalarOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [lower equal](burn_backend::ops::FloatTensorOps::float_lower_equal).
    /// Int => [lower equal](burn_backend::ops::IntTensorOps::int_lower_equal).
    LowerEqual(BinaryOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [lower equal elem](burn_backend::ops::FloatTensorOps::float_lower_equal_elem).
    /// Int => [lower equal elem](burn_backend::ops::IntTensorOps::int_lower_equal_elem).
    LowerEqualElem(ScalarOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [argmax](burn_backend::ops::FloatTensorOps::float_argmax).
    /// Int => [argmax](burn_backend::ops::IntTensorOps::int_argmax).
    ArgMax(ReduceDimOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [argtopk](burn_backend::ops::FloatTensorOps::float_argtopk).
    /// Int => [argtopk](burn_backend::ops::IntTensorOps::int_argtopk).
    ArgTopK(ReduceDimOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [topk](burn_backend::ops::FloatTensorOps::float_topk).
    /// Int => [topk](burn_backend::ops::IntTensorOps::int_topk).
    TopK(ReduceDimOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [topk with indices](burn_backend::ops::FloatTensorOps::float_topk_with_indices).
    /// Int => [topk with indices](burn_backend::ops::IntTensorOps::int_topk_with_indices).
    TopKWithIndices(TopKWithIndicesOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [argmin](burn_backend::ops::FloatTensorOps::float_argmin).
    /// Int => [argmin](burn_backend::ops::IntTensorOps::int_argmin).
    ArgMin(ReduceDimOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [max](burn_backend::ops::FloatTensorOps::float_max).
    /// Int => [max](burn_backend::ops::IntTensorOps::int_max).
    Max(ReduceOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [max dim with indices](burn_backend::ops::FloatTensorOps::float_max_dim_with_indices).
    /// Int => [max dim with indices](burn_backend::ops::IntTensorOps::int_max_dim_with_indices).
    MaxDimWithIndices(ReduceDimWithIndicesOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [min dim with indices](burn_backend::ops::FloatTensorOps::float_min_dim_with_indices).
    /// Int => [min dim with indices](burn_backend::ops::IntTensorOps::int_min_dim_with_indices).
    MinDimWithIndices(ReduceDimWithIndicesOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [min](burn_backend::ops::FloatTensorOps::float_min).
    /// Int => [min](burn_backend::ops::IntTensorOps::int_min).
    Min(ReduceOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [max dim](burn_backend::ops::FloatTensorOps::float_max_dim).
    /// Int => [max dim](burn_backend::ops::IntTensorOps::int_max_dim).
    MaxDim(ReduceDimOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [min dim](burn_backend::ops::FloatTensorOps::float_min_dim).
    /// Int => [min dim](burn_backend::ops::IntTensorOps::int_min_dim).
    MinDim(ReduceDimOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [max_abs](burn_backend::ops::FloatTensorOps::float_max_abs).
    /// Int => [max_abs](burn_backend::ops::IntTensorOps::int_max_abs).
    MaxAbs(ReduceOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [max_abs dim](burn_backend::ops::FloatTensorOps::float_max_abs_dim).
    /// Int => [max_abs dim](burn_backend::ops::IntTensorOps::int_max_abs_dim).
    MaxAbsDim(ReduceDimOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [clamp](burn_backend::ops::FloatTensorOps::float_clamp).
    /// Int => [clamp](burn_backend::ops::IntTensorOps::int_clamp).
    Clamp(ClampOpIr),
    /// Operation corresponding to:
    ///
    /// Int => [random](burn_backend::ops::IntTensorOps::int_random).
    IntRandom(RandomOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [powf](burn_backend::ops::FloatTensorOps::float_powi).
    /// Int => [powf](burn_backend::ops::IntTensorOps::int_powi).
    Powi(BinaryOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [powi_scalar](burn_backend::ops::FloatTensorOps::float_powi_scalar).
    /// Int => [powi_scalar](burn_backend::ops::IntTensorOps::int_powi_scalar).
    PowiScalar(ScalarOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [cumsum](burn_backend::ops::FloatTensorOps::float_cumsum).
    /// Int => [cumsum](burn_backend::ops::IntTensorOps::int_cumsum).
    CumSum(DimOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [cumprod](burn_backend::ops::FloatTensorOps::float_cumprod).
    /// Int => [cumprod](burn_backend::ops::IntTensorOps::int_cumprod).
    CumProd(DimOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [cummin](burn_backend::ops::FloatTensorOps::float_cummin).
    /// Int => [cummin](burn_backend::ops::IntTensorOps::int_cummin).
    CumMin(DimOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [cummax](burn_backend::ops::FloatTensorOps::float_cummax).
    /// Int => [cummax](burn_backend::ops::IntTensorOps::int_cummax).
    CumMax(DimOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [neg](burn_backend::ops::FloatTensorOps::float_neg).
    /// Int => [neg](burn_backend::ops::IntTensorOps::int_neg).
    Neg(UnaryOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [sign](burn_backend::ops::FloatTensorOps::float_sign).
    /// Int => [sign](burn_backend::ops::IntTensorOps::int_sign).
    Sign(UnaryOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [clamp_min](burn_backend::ops::FloatTensorOps::float_clamp_min).
    /// Int => [clamp_min](burn_backend::ops::IntTensorOps::int_clamp_min).
    ClampMin(ScalarOpIr),
    /// Operation corresponding to:
    ///
    /// Float => [clamp_max](burn_backend::ops::FloatTensorOps::float_clamp_max).
    /// Int => [clamp_max](burn_backend::ops::IntTensorOps::int_clamp_max).
    ClampMax(ScalarOpIr),
    /// Sort along a dim. Output shares the input's shape and dtype.
    Sort(SortOpIr),
    /// Sort along a dim, also returning the source indices.
    SortWithIndices(SortWithIndicesOpIr),
    /// Sort along a dim and return only the indices (i.e., the argsort).
    ///
    /// Shares [`SortOpIr`] with [`Sort`](Self::Sort); only the output dtype differs.
    ArgSort(SortOpIr),
}

/// Operation intermediate representation specific to an int tensor.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub enum IntOperationIr {
    /// Operation corresponding to [into float](burn_backend::ops::IntTensorOps::int_into_float).
    IntoFloat(CastOpIr),
    /// Operation corresponding to:
    ///
    /// Int => [bitwise and](burn_backend::ops::IntTensorOps::bitwise_and).
    BitwiseAnd(BinaryOpIr),
    /// Operation corresponding to:
    ///
    /// Int => [bitwise and scalar](burn_backend::ops::IntTensorOps::bitwise_and_scalar).
    BitwiseAndScalar(ScalarOpIr),
    /// Operation corresponding to:
    ///
    /// Int => [bitwise or](burn_backend::ops::IntTensorOps::bitwise_or).
    BitwiseOr(BinaryOpIr),
    /// Operation corresponding to:
    ///
    /// Int => [bitwise or scalar](burn_backend::ops::IntTensorOps::bitwise_or_scalar).
    BitwiseOrScalar(ScalarOpIr),
    /// Operation corresponding to:
    ///
    /// Int => [bitwise xor](burn_backend::ops::IntTensorOps::bitwise_xor).
    BitwiseXor(BinaryOpIr),
    /// Operation corresponding to:
    ///
    /// Int => [bitwise xor scalar](burn_backend::ops::IntTensorOps::bitwise_xor_scalar).
    BitwiseXorScalar(ScalarOpIr),
    /// Operation corresponding to:
    ///
    /// Int => [bitwise not](burn_backend::ops::IntTensorOps::bitwise_not).
    BitwiseNot(UnaryOpIr),
    /// Operation corresponding to:
    ///
    /// Int => [bitwise left shift](burn_backend::ops::IntTensorOps::bitwise_left_shift).
    BitwiseLeftShift(BinaryOpIr),
    /// Operation corresponding to:
    ///
    /// Int => [bitwise left shift scalar](burn_backend::ops::IntTensorOps::bitwise_left_shift_scalar).
    BitwiseLeftShiftScalar(ScalarOpIr),
    /// Operation corresponding to:
    ///
    /// Int => [bitwise right shift](burn_backend::ops::IntTensorOps::bitwise_right_shift).
    BitwiseRightShift(BinaryOpIr),
    /// Operation corresponding to:
    ///
    /// Int => [bitwise right shift scalar](burn_backend::ops::IntTensorOps::bitwise_right_shift_scalar).
    BitwiseRightShiftScalar(ScalarOpIr),
    /// Operation corresponding to [matmul](burn_backend::ops::IntTensorOps::int_matmul).
    Matmul(MatmulOpIr),
}

/// Operation intermediate representation specific to a bool tensor.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub enum BoolOperationIr {
    /// Operation corresponding to [into float](burn_backend::ops::BoolTensorOps::bool_into_float).
    IntoFloat(CastOpIr),
    /// Operation corresponding to [into int](burn_backend::ops::BoolTensorOps::bool_into_int).
    IntoInt(CastOpIr),
    /// Operation corresponding to [not](burn_backend::ops::BoolTensorOps::bool_not).
    Not(UnaryOpIr),
    /// Operation corresponding to [and](burn_backend::ops::BoolTensorOps::bool_and).
    And(BinaryOpIr),
    /// Operation corresponding to [or](burn_backend::ops::BoolTensorOps::bool_or).
    Or(BinaryOpIr),
    /// Operation corresponding to [xor](burn_backend::ops::BoolTensorOps::bool_xor).
    Xor(BinaryOpIr),
}

/// Operations that can be done on distributed tensors.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(clippy::large_enum_variant)]
pub enum DistributedOperationIr {
    /// Operation corresponding to:
    /// [all_reduce](burn_backend::distributed::DistributedOps::all_reduce).
    AllReduce(AllReduceOpIr),
    /// Resolve the pending collective operations on the executing device. Corresponds to
    /// [sync_collective](burn_backend::distributed::DistributedOps::sync_collective).
    ///
    /// Fire-and-forget and payload-free: it syncs whichever device the interpreter is bound to.
    /// Modeled as an operation (not a side-channel call) so it travels the normal op stream
    /// alongside [`AllReduce`](DistributedOperationIr::AllReduce) and is ordered against it.
    SyncCollective,
}

/// Swap dim operation intermediate representation.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct SwapDimsOpIr {
    /// Input tensor intermediate representation.
    pub input: TensorIr,
    /// Output tensor intermediate representation.
    pub out: TensorIr,
    /// The first dim to swap.
    pub dim1: usize,
    /// The second dim to swap.
    pub dim2: usize,
}

/// Permute operation intermediate representation.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct PermuteOpIr {
    /// Input tensor intermediate representation.
    pub input: TensorIr,
    /// Output tensor intermediate representation.
    pub out: TensorIr,
    /// The new order of the dimensions.
    pub axes: Vec<usize>,
}

/// Shape operation intermediate representation.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct ShapeOpIr {
    /// Input tensor intermediate representation.
    pub input: TensorIr,
    /// Output tensor intermediate representation with the new shape.
    pub out: TensorIr,
}

/// Unfold operation intermediate representation.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct UnfoldOpIr {
    /// Input tensor intermediate representation.
    pub input: TensorIr,
    /// Output tensor intermediate representation.
    pub out: TensorIr,

    /// The selected dim.
    pub dim: usize,
    /// The window size.
    pub size: usize,
    /// The window step along dim.
    pub step: usize,
}

/// Flip operation intermediate representation.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct FlipOpIr {
    /// Input tensor intermediate representation.
    pub input: TensorIr,
    /// Output tensor intermediate representation.
    pub out: TensorIr,
    /// The dimensions to flip.
    pub axes: Vec<usize>,
}

#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct RandomOpIr {
    pub out: TensorIr,
    pub distribution: Distribution,
}

/// Creation operation intermediate representation.
/// As opposed to [InitOperationIr], creation operations are lazy initialized.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct CreationOpIr {
    /// Output tensor intermediate representation.
    pub out: TensorIr,
}

/// Full operation intermediate representation.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct FullOpIr {
    /// Output tensor intermediate representation.
    pub out: TensorIr,
    /// Fill value.
    pub value: ScalarIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
/// Declares a tensor has been initialized.
///
/// It is necessary to register for proper orphan detection and avoid memory leak.
pub struct InitOperationIr {
    /// The initialized tensor.
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct BinaryOpIr {
    pub lhs: TensorIr,
    pub rhs: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct MatmulOpIr {
    pub lhs: TensorIr,
    pub rhs: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct CrossOpIr {
    pub lhs: TensorIr,
    pub rhs: TensorIr,
    pub out: TensorIr,
    pub dim: usize,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct UnaryOpIr {
    pub input: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct ScalarOpIr {
    pub lhs: TensorIr,
    // TODO: Make that an enum with `Value` and `Id` variants for relative/global
    // conversion.
    pub rhs: ScalarIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, PartialEq, Serialize, Deserialize, Hash)]
#[allow(missing_docs)]
pub struct ReduceOpIr {
    pub input: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, PartialEq, Serialize, Deserialize, Hash)]
#[allow(missing_docs)]
pub struct ReduceDimOpIr {
    pub input: TensorIr,
    pub out: TensorIr,
    pub axis: usize,
    pub accumulator_len: usize,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct CastOpIr {
    pub input: TensorIr,
    pub out: TensorIr,
}

/// IR for operations that operate along a dimension without reducing it.
/// Unlike `ReduceDimOpIr`, the output shape is the same as the input shape.
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize, Hash)]
#[allow(missing_docs)]
pub struct DimOpIr {
    pub input: TensorIr,
    pub out: TensorIr,
    pub axis: usize,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct GatherOpIr {
    pub tensor: TensorIr,
    pub dim: usize,
    pub indices: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct ScatterOpIr {
    pub tensor: TensorIr,
    pub dim: usize,
    pub indices: TensorIr,
    pub value: TensorIr,
    pub update: IndexingUpdateOp,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct ScatterNdOpIr {
    pub data: TensorIr,
    pub indices: TensorIr,
    pub values: TensorIr,
    pub reduction: IndexingUpdateOp,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct GatherNdOpIr {
    pub data: TensorIr,
    pub indices: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct SelectOpIr {
    pub tensor: TensorIr,
    pub dim: usize,
    pub indices: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct SelectAssignOpIr {
    pub tensor: TensorIr,
    pub dim: usize,
    pub indices: TensorIr,
    pub value: TensorIr,
    pub update: IndexingUpdateOp,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct SliceOpIr {
    pub tensor: TensorIr,
    pub ranges: Vec<Slice>,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct SliceAssignOpIr {
    pub tensor: TensorIr,
    pub ranges: Vec<burn_backend::Slice>,
    pub value: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct MaskWhereOpIr {
    pub tensor: TensorIr,
    pub mask: TensorIr,
    pub value: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct MaskFillOpIr {
    pub tensor: TensorIr,
    pub mask: TensorIr,
    pub value: ScalarIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct ClampOpIr {
    pub tensor: TensorIr,
    pub min: ScalarIr,
    pub max: ScalarIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct RepeatDimOpIr {
    pub tensor: TensorIr,
    pub dim: usize,
    pub times: usize,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct CatOpIr {
    pub tensors: Vec<TensorIr>,
    pub dim: usize,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct AllReduceOpIr {
    pub tensor: TensorIr,
    pub out: TensorIr,
    /// How to reduce the values across the participating devices.
    pub op: burn_backend::distributed::ReduceOperation,
    /// The devices participating in the collective operation.
    pub device_ids: Vec<DeviceIdIr>,
}

/// Serializable representation of a [device id](burn_backend::DeviceId).
///
/// The intermediate representation is part of the wire protocol (e.g. the remote backend), so it
/// cannot store `burn_backend::DeviceId` directly since that type is not serializable.
#[derive(Clone, Copy, Debug, Hash, PartialEq, Eq, Serialize, Deserialize)]
pub struct DeviceIdIr {
    /// Identifies the type of the device.
    pub type_id: u16,
    /// Identifies the device number.
    pub index_id: u16,
}

impl From<burn_backend::DeviceId> for DeviceIdIr {
    fn from(value: burn_backend::DeviceId) -> Self {
        Self {
            type_id: value.type_id,
            index_id: value.index_id,
        }
    }
}

impl From<DeviceIdIr> for burn_backend::DeviceId {
    fn from(value: DeviceIdIr) -> Self {
        burn_backend::DeviceId::new(value.type_id, value.index_id)
    }
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct ReduceDimWithIndicesOpIr {
    pub tensor: TensorIr,
    pub dim: usize,
    pub out: TensorIr,
    pub out_indices: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
/// Like [`ReduceDimWithIndicesOpIr`], but for a top-k: the reduced axis keeps `k` entries
/// instead of collapsing to 1, so `k` has to be carried explicitly.
pub struct TopKWithIndicesOpIr {
    pub tensor: TensorIr,
    pub dim: usize,
    pub k: usize,
    pub out: TensorIr,
    pub out_indices: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct EmbeddingOpIr {
    pub weights: TensorIr,
    pub indices: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct EmbeddingBackwardOpIr {
    pub weights: TensorIr,
    pub out_grad: TensorIr,
    pub indices: TensorIr,
    pub out: TensorIr,
}

/// Batch normalization using explicitly supplied channel statistics.
///
/// This operation neither calculates nor updates the supplied statistics.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct BatchNormOpIr {
    pub x: TensorIr,
    pub gamma: TensorIr,
    pub beta: TensorIr,
    pub mean: TensorIr,
    pub variance: TensorIr,
    pub epsilon: ScalarIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct LinearOpIr {
    pub x: TensorIr,
    pub weight: TensorIr,
    pub bias: Option<TensorIr>,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct LinearXBackwardOpIr {
    pub weight: TensorIr,
    pub output_grad: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct LinearWeightBackwardOpIr {
    pub x: TensorIr,
    pub output_grad: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct LinearBiasBackwardOpIr {
    pub output_grad: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct Conv1dOpIr {
    pub x: TensorIr,
    pub weight: TensorIr,
    pub bias: Option<TensorIr>,
    pub options: Conv1dOptionsIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct Conv1dXBackwardOpIr {
    pub x: TensorIr,
    pub weight: TensorIr,
    pub output_grad: TensorIr,
    pub options: Conv1dOptionsIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct Conv1dWeightBackwardOpIr {
    pub x: TensorIr,
    pub weight: TensorIr,
    pub output_grad: TensorIr,
    pub options: Conv1dOptionsIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct Conv1dBiasBackwardOpIr {
    pub x: TensorIr,
    pub bias: TensorIr,
    pub output_grad: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct Conv2dOpIr {
    pub x: TensorIr,
    pub weight: TensorIr,
    pub bias: Option<TensorIr>,
    pub options: Conv2dOptionsIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct Conv2dXBackwardOpIr {
    pub x: TensorIr,
    pub weight: TensorIr,
    pub output_grad: TensorIr,
    pub options: Conv2dOptionsIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct Conv2dWeightBackwardOpIr {
    pub x: TensorIr,
    pub weight: TensorIr,
    pub output_grad: TensorIr,
    pub options: Conv2dOptionsIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct Conv2dBiasBackwardOpIr {
    pub x: TensorIr,
    pub bias: TensorIr,
    pub output_grad: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct DeformConv2dOpIr {
    pub x: TensorIr,
    pub offset: TensorIr,
    pub weight: TensorIr,
    pub mask: Option<TensorIr>,
    pub bias: Option<TensorIr>,
    pub options: DeformableConv2dOptionsIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct DeformConv2dBackwardOpIr {
    pub x: TensorIr,
    pub offset: TensorIr,
    pub weight: TensorIr,
    pub mask: Option<TensorIr>,
    pub bias: Option<TensorIr>,
    pub out_grad: TensorIr,
    pub options: DeformableConv2dOptionsIr,
    pub input_grad: TensorIr,
    pub offset_grad: TensorIr,
    pub weight_grad: TensorIr,
    pub mask_grad: Option<TensorIr>,
    pub bias_grad: Option<TensorIr>,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct Conv3dOpIr {
    pub x: TensorIr,
    pub weight: TensorIr,
    pub bias: Option<TensorIr>,
    pub options: Conv3dOptionsIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct Conv3dXBackwardOpIr {
    pub x: TensorIr,
    pub weight: TensorIr,
    pub output_grad: TensorIr,
    pub options: Conv3dOptionsIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct Conv3dWeightBackwardOpIr {
    pub x: TensorIr,
    pub weight: TensorIr,
    pub output_grad: TensorIr,
    pub options: Conv3dOptionsIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct Conv3dBiasBackwardOpIr {
    pub x: TensorIr,
    pub bias: TensorIr,
    pub output_grad: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct ConvTranspose1dOpIr {
    pub x: TensorIr,
    pub weight: TensorIr,
    pub bias: Option<TensorIr>,
    pub options: ConvTranspose1dOptionsIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct ConvTranspose2dOpIr {
    pub x: TensorIr,
    pub weight: TensorIr,
    pub bias: Option<TensorIr>,
    pub options: ConvTranspose2dOptionsIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct ConvTranspose3dOpIr {
    pub x: TensorIr,
    pub weight: TensorIr,
    pub bias: Option<TensorIr>,
    pub options: ConvTranspose3dOptionsIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct Conv1dOptionsIr {
    pub stride: [usize; 1],
    pub padding: [usize; 1],
    pub dilation: [usize; 1],
    pub groups: usize,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct Conv2dOptionsIr {
    pub stride: [usize; 2],
    pub padding: [usize; 2],
    pub dilation: [usize; 2],
    pub groups: usize,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct DeformableConv2dOptionsIr {
    pub stride: [usize; 2],
    pub padding: [usize; 2],
    pub dilation: [usize; 2],
    pub weight_groups: usize,
    pub offset_groups: usize,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct Conv3dOptionsIr {
    pub stride: [usize; 3],
    pub padding: [usize; 3],
    pub dilation: [usize; 3],
    pub groups: usize,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct ConvTranspose1dOptionsIr {
    pub stride: [usize; 1],
    pub padding: [usize; 1],
    pub padding_out: [usize; 1],
    pub dilation: [usize; 1],
    pub groups: usize,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct ConvTranspose2dOptionsIr {
    pub stride: [usize; 2],
    pub padding: [usize; 2],
    pub padding_out: [usize; 2],
    pub dilation: [usize; 2],
    pub groups: usize,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct ConvTranspose3dOptionsIr {
    pub stride: [usize; 3],
    pub padding: [usize; 3],
    pub padding_out: [usize; 3],
    pub dilation: [usize; 3],
    pub groups: usize,
}

/// Quantization parameters intermediate representation.
#[derive(Debug, Clone, Hash, PartialEq, Eq, Serialize, Deserialize)]
pub struct QuantizationParametersIr {
    /// The scaling factor, one per block or a single one for a per-tensor level.
    pub scales: TensorIr,
    /// The per-tensor scale that [`scales`](Self::scales) are expressed relative to, for a
    /// two-level scheme.
    pub global: Option<TensorIr>,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct QuantizeOpIr {
    pub tensor: TensorIr,
    pub qparams: QuantizationParametersIr,
    pub scheme: QuantScheme,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct DequantizeOpIr {
    pub input: TensorIr,
    pub out: TensorIr,
}

impl From<ConvOptions<1>> for Conv1dOptionsIr {
    fn from(value: ConvOptions<1>) -> Self {
        Self {
            stride: value.stride,
            padding: value.padding,
            dilation: value.dilation,
            groups: value.groups,
        }
    }
}

impl From<ConvOptions<2>> for Conv2dOptionsIr {
    fn from(value: ConvOptions<2>) -> Self {
        Self {
            stride: value.stride,
            padding: value.padding,
            dilation: value.dilation,
            groups: value.groups,
        }
    }
}

impl From<ConvOptions<3>> for Conv3dOptionsIr {
    fn from(value: ConvOptions<3>) -> Self {
        Self {
            stride: value.stride,
            padding: value.padding,
            dilation: value.dilation,
            groups: value.groups,
        }
    }
}

impl From<DeformConvOptions<2>> for DeformableConv2dOptionsIr {
    fn from(value: DeformConvOptions<2>) -> Self {
        Self {
            stride: value.stride,
            padding: value.padding,
            dilation: value.dilation,
            weight_groups: value.weight_groups,
            offset_groups: value.offset_groups,
        }
    }
}

impl From<ConvTransposeOptions<1>> for ConvTranspose1dOptionsIr {
    fn from(value: ConvTransposeOptions<1>) -> Self {
        Self {
            stride: value.stride,
            padding: value.padding,
            padding_out: value.padding_out,
            dilation: value.dilation,
            groups: value.groups,
        }
    }
}

impl From<ConvTransposeOptions<2>> for ConvTranspose2dOptionsIr {
    fn from(value: ConvTransposeOptions<2>) -> Self {
        Self {
            stride: value.stride,
            padding: value.padding,
            padding_out: value.padding_out,
            dilation: value.dilation,
            groups: value.groups,
        }
    }
}

impl From<ConvTransposeOptions<3>> for ConvTranspose3dOptionsIr {
    fn from(value: ConvTransposeOptions<3>) -> Self {
        Self {
            stride: value.stride,
            padding: value.padding,
            padding_out: value.padding_out,
            dilation: value.dilation,
            groups: value.groups,
        }
    }
}

impl From<Conv1dOptionsIr> for ConvOptions<1> {
    fn from(val: Conv1dOptionsIr) -> Self {
        ConvOptions {
            stride: val.stride,
            padding: val.padding,
            dilation: val.dilation,
            groups: val.groups,
        }
    }
}

impl From<Conv2dOptionsIr> for ConvOptions<2> {
    fn from(val: Conv2dOptionsIr) -> Self {
        ConvOptions {
            stride: val.stride,
            padding: val.padding,
            dilation: val.dilation,
            groups: val.groups,
        }
    }
}

impl From<Conv3dOptionsIr> for ConvOptions<3> {
    fn from(val: Conv3dOptionsIr) -> Self {
        ConvOptions {
            stride: val.stride,
            padding: val.padding,
            dilation: val.dilation,
            groups: val.groups,
        }
    }
}

impl From<DeformableConv2dOptionsIr> for DeformConvOptions<2> {
    fn from(value: DeformableConv2dOptionsIr) -> Self {
        DeformConvOptions {
            stride: value.stride,
            padding: value.padding,
            dilation: value.dilation,
            weight_groups: value.weight_groups,
            offset_groups: value.offset_groups,
        }
    }
}

impl From<ConvTranspose1dOptionsIr> for ConvTransposeOptions<1> {
    fn from(val: ConvTranspose1dOptionsIr) -> Self {
        ConvTransposeOptions {
            stride: val.stride,
            padding: val.padding,
            padding_out: val.padding_out,
            dilation: val.dilation,
            groups: val.groups,
        }
    }
}

impl From<ConvTranspose2dOptionsIr> for ConvTransposeOptions<2> {
    fn from(val: ConvTranspose2dOptionsIr) -> Self {
        ConvTransposeOptions {
            stride: val.stride,
            padding: val.padding,
            padding_out: val.padding_out,
            dilation: val.dilation,
            groups: val.groups,
        }
    }
}

impl From<ConvTranspose3dOptionsIr> for ConvTransposeOptions<3> {
    fn from(val: ConvTranspose3dOptionsIr) -> Self {
        ConvTransposeOptions {
            stride: val.stride,
            padding: val.padding,
            padding_out: val.padding_out,
            dilation: val.dilation,
            groups: val.groups,
        }
    }
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct AvgPool1dOpIr {
    pub x: TensorIr,
    pub kernel_size: usize,
    pub stride: usize,
    pub padding: usize,
    pub count_include_pad: bool,
    pub ceil_mode: bool,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct AvgPool2dOpIr {
    pub x: TensorIr,
    pub kernel_size: [usize; 2],
    pub stride: [usize; 2],
    pub padding: [usize; 2],
    pub count_include_pad: bool,
    pub ceil_mode: bool,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct AvgPool1dBackwardOpIr {
    pub x: TensorIr,
    pub grad: TensorIr,
    pub kernel_size: usize,
    pub stride: usize,
    pub padding: usize,
    pub count_include_pad: bool,
    pub ceil_mode: bool,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct AvgPool2dBackwardOpIr {
    pub x: TensorIr,
    pub grad: TensorIr,
    pub kernel_size: [usize; 2],
    pub stride: [usize; 2],
    pub padding: [usize; 2],
    pub count_include_pad: bool,
    pub ceil_mode: bool,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct AdaptiveAvgPool1dOpIr {
    pub x: TensorIr,
    pub output_size: usize,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct AdaptiveAvgPool2dOpIr {
    pub x: TensorIr,
    pub output_size: [usize; 2],
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct AdaptiveAvgPool1dBackwardOpIr {
    pub x: TensorIr,
    pub grad: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct AdaptiveAvgPool2dBackwardOpIr {
    pub x: TensorIr,
    pub grad: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct AdaptiveAvgPool3dOpIr {
    pub x: TensorIr,
    pub output_size: [usize; 3],
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct AdaptiveAvgPool3dBackwardOpIr {
    pub x: TensorIr,
    pub grad: TensorIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct MaxPool1dOpIr {
    pub x: TensorIr,
    pub kernel_size: usize,
    pub stride: usize,
    pub padding: usize,
    pub dilation: usize,
    pub ceil_mode: bool,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct MaxPool1dWithIndicesOpIr {
    pub x: TensorIr,
    pub kernel_size: usize,
    pub stride: usize,
    pub padding: usize,
    pub dilation: usize,
    pub ceil_mode: bool,
    pub out: TensorIr,
    pub out_indices: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct MaxPool1dWithIndicesBackwardOpIr {
    pub x: TensorIr,
    pub grad: TensorIr,
    pub indices: TensorIr,
    pub kernel_size: usize,
    pub stride: usize,
    pub padding: usize,
    pub dilation: usize,
    pub ceil_mode: bool,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct MaxPool2dOpIr {
    pub x: TensorIr,
    pub kernel_size: [usize; 2],
    pub stride: [usize; 2],
    pub padding: [usize; 2],
    pub dilation: [usize; 2],
    pub ceil_mode: bool,
    pub out: TensorIr,
}

#[allow(missing_docs)]
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct MaxPool2dWithIndicesOpIr {
    pub x: TensorIr,
    pub kernel_size: [usize; 2],
    pub stride: [usize; 2],
    pub padding: [usize; 2],
    pub dilation: [usize; 2],
    pub ceil_mode: bool,
    pub out: TensorIr,
    pub out_indices: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct MaxPool2dWithIndicesBackwardOpIr {
    pub x: TensorIr,
    pub grad: TensorIr,
    pub indices: TensorIr,
    pub kernel_size: [usize; 2],
    pub stride: [usize; 2],
    pub padding: [usize; 2],
    pub dilation: [usize; 2],
    pub ceil_mode: bool,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub enum InterpolateModeIr {
    Nearest,
    NearestExact,
    Bilinear,
    Bicubic,
    Lanczos3,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct InterpolateOptionsIr {
    pub mode: InterpolateModeIr,
    pub align_corners: bool,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct InterpolateOpIr {
    pub x: TensorIr,
    pub output_size: [usize; 2],
    pub options: InterpolateOptionsIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct RfftOpIr {
    pub signal: TensorIr,
    pub dim: usize,
    pub n: Option<usize>,
    pub out_re: TensorIr,
    pub out_im: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct IRfftOpIr {
    pub input_re: TensorIr,
    pub input_im: TensorIr,
    pub dim: usize,
    pub n: Option<usize>,
    pub out_signal: TensorIr,
}

#[allow(missing_docs)]
impl RfftOpIr {
    pub fn create<F>(signal: TensorIr, dim: usize, n: Option<usize>, mut new_id: F) -> Self
    where
        F: FnMut() -> crate::TensorId,
    {
        // `n` is required to be a power of two at the public API boundary, so
        // the output has `n / 2 + 1` bins (matching scipy/torch for pow2 n).
        let mut shape = signal.shape.clone();
        let fft_len = n.unwrap_or(shape[dim]);
        shape[dim] = fft_len / 2 + 1;
        let dtype = signal.dtype;

        Self {
            signal,
            dim,
            n,
            out_re: TensorIr::uninit(new_id(), shape.clone(), dtype),
            out_im: TensorIr::uninit(new_id(), shape, dtype),
        }
    }
}

#[allow(missing_docs)]
impl IRfftOpIr {
    pub fn create<F>(
        input_re: TensorIr,
        input_im: TensorIr,
        dim: usize,
        n: Option<usize>,
        mut new_id: F,
    ) -> Self
    where
        F: FnMut() -> crate::TensorId,
    {
        debug_assert!(
            input_re.shape[dim] >= 1,
            "IRfftOpIr: input spectrum dimension must be >= 1"
        );
        debug_assert!(
            !matches!(n, Some(0)),
            "IRfftOpIr: n must be >= 1 when specified"
        );
        let mut shape = input_re.shape.clone();
        shape[dim] = n.unwrap_or((shape[dim] - 1) * 2);
        let dtype = input_re.dtype;

        Self {
            input_re,
            input_im,
            dim,
            n,
            out_signal: TensorIr::uninit(new_id(), shape, dtype),
        }
    }
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct AttentionOptionsIr {
    pub scale: Option<ScalarIr>,
    pub softcap: Option<ScalarIr>,
    pub is_causal: bool,
}

impl From<AttentionOptionsIr> for AttentionModuleOptions {
    fn from(ir: AttentionOptionsIr) -> Self {
        AttentionModuleOptions {
            scale: ir.scale.map(|s| s.elem()),
            softcap: ir.softcap.map(|s| s.elem()),
            is_causal: ir.is_causal,
        }
    }
}

impl From<AttentionModuleOptions> for AttentionOptionsIr {
    fn from(ir: AttentionModuleOptions) -> Self {
        AttentionOptionsIr {
            scale: ir.scale.map(ScalarIr::Float),
            softcap: ir.softcap.map(ScalarIr::Float),
            is_causal: ir.is_causal,
        }
    }
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct AttentionOpIr {
    pub query: TensorIr,
    pub key: TensorIr,
    pub value: TensorIr,
    pub mask: Option<TensorIr>,
    pub attn_bias: Option<TensorIr>,
    pub options: AttentionOptionsIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct CtcLossOpIr {
    pub log_probs: TensorIr,
    pub targets: TensorIr,
    pub input_lengths: TensorIr,
    pub target_lengths: TensorIr,
    pub blank: usize,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct CtcLossBackwardOpIr {
    pub log_probs: TensorIr,
    pub targets: TensorIr,
    pub input_lengths: TensorIr,
    pub target_lengths: TensorIr,
    pub grad_loss: TensorIr,
    pub blank: usize,
    pub out: TensorIr,
}

impl From<InterpolateModeIr> for InterpolateMode {
    fn from(val: InterpolateModeIr) -> Self {
        match val {
            InterpolateModeIr::Nearest => Self::Nearest,
            InterpolateModeIr::NearestExact => Self::NearestExact,
            InterpolateModeIr::Bilinear => Self::Bilinear,
            InterpolateModeIr::Bicubic => Self::Bicubic,
            InterpolateModeIr::Lanczos3 => Self::Lanczos3,
        }
    }
}

impl From<InterpolateOptionsIr> for InterpolateOptions {
    fn from(val: InterpolateOptionsIr) -> Self {
        Self::new(val.mode.into()).with_align_corners(val.align_corners)
    }
}

impl From<InterpolateMode> for InterpolateModeIr {
    fn from(val: InterpolateMode) -> Self {
        match val {
            InterpolateMode::Nearest => Self::Nearest,
            InterpolateMode::NearestExact => Self::NearestExact,
            InterpolateMode::Bilinear => Self::Bilinear,
            InterpolateMode::Bicubic => Self::Bicubic,
            InterpolateMode::Lanczos3 => Self::Lanczos3,
        }
    }
}

impl From<InterpolateOptions> for InterpolateOptionsIr {
    fn from(val: InterpolateOptions) -> Self {
        Self {
            mode: val.mode.into(),
            align_corners: val.align_corners,
        }
    }
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct InterpolateBackwardOpIr {
    pub x: TensorIr,
    pub grad: TensorIr,
    pub output_size: [usize; 2],
    pub options: InterpolateOptionsIr,
    pub out: TensorIr,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub enum GridSamplePaddingModeIr {
    Zeros,
    Border,
    Reflection,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct GridSampleOptionsIr {
    pub mode: InterpolateModeIr,
    pub padding_mode: GridSamplePaddingModeIr,
    pub align_corners: bool,
}

#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
#[allow(missing_docs)]
pub struct GridSample2dOpIr {
    pub tensor: TensorIr,
    pub grid: TensorIr,
    pub options: GridSampleOptionsIr,
    pub out: TensorIr,
}

impl From<GridSamplePaddingModeIr> for GridSamplePaddingMode {
    fn from(val: GridSamplePaddingModeIr) -> Self {
        match val {
            GridSamplePaddingModeIr::Zeros => Self::Zeros,
            GridSamplePaddingModeIr::Border => Self::Border,
            GridSamplePaddingModeIr::Reflection => Self::Reflection,
        }
    }
}

impl From<GridSamplePaddingMode> for GridSamplePaddingModeIr {
    fn from(val: GridSamplePaddingMode) -> Self {
        match val {
            GridSamplePaddingMode::Zeros => Self::Zeros,
            GridSamplePaddingMode::Border => Self::Border,
            GridSamplePaddingMode::Reflection => Self::Reflection,
        }
    }
}

impl From<GridSampleOptionsIr> for GridSampleOptions {
    fn from(val: GridSampleOptionsIr) -> Self {
        Self {
            mode: val.mode.into(),
            padding_mode: val.padding_mode.into(),
            align_corners: val.align_corners,
        }
    }
}

impl From<GridSampleOptions> for GridSampleOptionsIr {
    fn from(val: GridSampleOptions) -> Self {
        Self {
            mode: val.mode.into(),
            padding_mode: val.padding_mode.into(),
            align_corners: val.align_corners,
        }
    }
}

impl OperationIr {
    /// Get all input [tensors](TensorIr) involved with the current operation.
    pub fn inputs(&self) -> impl Iterator<Item = &TensorIr> {
        match self {
            OperationIr::BaseFloat(repr) => repr.inputs(),
            OperationIr::BaseInt(repr) => repr.inputs(),
            OperationIr::BaseBool(repr) => repr.inputs(),
            OperationIr::NumericFloat(_dtype, repr) => repr.inputs(),
            OperationIr::NumericInt(_dtype, repr) => repr.inputs(),
            OperationIr::Bool(repr) => repr.inputs(),
            OperationIr::Int(repr) => repr.inputs(),
            OperationIr::Float(_dtype, repr) => repr.inputs(),
            OperationIr::Module(repr) => repr.inputs(),
            OperationIr::Init(repr) => repr.inputs(),
            OperationIr::Custom(repr) => repr.inputs(),
            OperationIr::Drop(repr) => Box::new([repr].into_iter()),
            OperationIr::Distributed(repr) => repr.inputs(),
            OperationIr::Activation(repr) => repr.inputs(),
        }
    }

    /// Get all output [tensors](TensorIr) involved with the current operation.
    pub fn outputs(&self) -> impl Iterator<Item = &TensorIr> {
        match self {
            OperationIr::BaseFloat(repr) => repr.outputs(),
            OperationIr::BaseInt(repr) => repr.outputs(),
            OperationIr::BaseBool(repr) => repr.outputs(),
            OperationIr::NumericFloat(_dtype, repr) => repr.outputs(),
            OperationIr::NumericInt(_dtype, repr) => repr.outputs(),
            OperationIr::Bool(repr) => repr.outputs(),
            OperationIr::Int(repr) => repr.outputs(),
            OperationIr::Float(_dtype, repr) => repr.outputs(),
            OperationIr::Module(repr) => repr.outputs(),
            OperationIr::Init(repr) => repr.outputs(),
            OperationIr::Custom(repr) => repr.outputs(),
            OperationIr::Drop(_repr) => Box::new([].into_iter()),
            OperationIr::Distributed(repr) => repr.outputs(),
            OperationIr::Activation(repr) => repr.outputs(),
        }
    }

    /// Get all [tensor](TensorIr) involved with the current operation.
    pub fn nodes(&self) -> Vec<&TensorIr> {
        self.inputs().chain(self.outputs()).collect()
    }

    /// Set the given nodes that are [read write](super::TensorStatus::ReadWrite) to
    /// [read only](super::TensorStatus::ReadOnly) in the current operation.
    ///
    /// Returns the tensor that were updated with their original representation.
    pub fn mark_read_only(&mut self, nodes: &[TensorId]) -> Vec<TensorIr> {
        match self {
            OperationIr::BaseFloat(repr) => repr.mark_read_only(nodes),
            OperationIr::BaseInt(repr) => repr.mark_read_only(nodes),
            OperationIr::BaseBool(repr) => repr.mark_read_only(nodes),
            OperationIr::NumericFloat(_dtype, repr) => repr.mark_read_only(nodes),
            OperationIr::NumericInt(_dtype, repr) => repr.mark_read_only(nodes),
            OperationIr::Bool(repr) => repr.mark_read_only(nodes),
            OperationIr::Int(repr) => repr.mark_read_only(nodes),
            OperationIr::Float(_dtype, repr) => repr.mark_read_only(nodes),
            OperationIr::Module(repr) => repr.mark_read_only(nodes),
            OperationIr::Init(_) => Vec::new(),
            OperationIr::Drop(repr) => {
                let mut output = Vec::new();
                repr.mark_read_only(nodes, &mut output);
                output
            }
            OperationIr::Custom(repr) => {
                let mut output = Vec::new();

                for input in repr.inputs.iter_mut() {
                    input.mark_read_only(nodes, &mut output);
                }

                output
            }
            OperationIr::Distributed(repr) => repr.mark_read_only(nodes),
            OperationIr::Activation(repr) => repr.mark_read_only(nodes),
        }
    }

    /// Visit every mutable component of this operation in place.
    ///
    /// For each variant, visits its [tensors](TensorIr) (in the same order as
    /// [`inputs`](Self::inputs) then [`outputs`](Self::outputs) combined), then its
    /// [scalars](ScalarIr), then its slice ranges. Only `Slice`/`SliceAssign` (under the `Base*`
    /// variants) carry ranges; every other category visits no ranges.
    pub fn visit_mut(&mut self, v: &mut impl IrVisitorMut) {
        match self {
            OperationIr::BaseFloat(repr) => repr.visit_mut(v),
            OperationIr::BaseInt(repr) => repr.visit_mut(v),
            OperationIr::BaseBool(repr) => repr.visit_mut(v),
            OperationIr::NumericFloat(_dtype, repr) => repr.visit_mut(v),
            OperationIr::NumericInt(_dtype, repr) => repr.visit_mut(v),
            OperationIr::Bool(repr) => repr.visit_mut(v),
            OperationIr::Int(repr) => repr.visit_mut(v),
            OperationIr::Float(_dtype, repr) => repr.visit_mut(v),
            OperationIr::Module(repr) => repr.visit_mut(v),
            OperationIr::Init(repr) => repr.visit_mut(v),
            OperationIr::Custom(repr) => repr.visit_mut(v),
            OperationIr::Drop(repr) => v.visit_tensor_mut(repr),
            OperationIr::Distributed(repr) => repr.visit_mut(v),
            OperationIr::Activation(repr) => repr.visit_mut(v),
        }
    }
}

impl BaseOperationIr {
    fn inputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        match self {
            BaseOperationIr::Reshape(repr) => Box::new([&repr.input].into_iter()),
            BaseOperationIr::SwapDims(repr) => Box::new([&repr.input].into_iter()),
            BaseOperationIr::Permute(repr) => Box::new([&repr.input].into_iter()),
            BaseOperationIr::Expand(repr) => Box::new([&repr.input].into_iter()),
            BaseOperationIr::Flip(repr) => Box::new([&repr.input].into_iter()),
            BaseOperationIr::Slice(repr) => Box::new([&repr.tensor].into_iter()),
            BaseOperationIr::SliceAssign(repr) => Box::new([&repr.tensor, &repr.value].into_iter()),
            BaseOperationIr::Gather(repr) => Box::new([&repr.tensor, &repr.indices].into_iter()),
            BaseOperationIr::Scatter(repr) => {
                Box::new([&repr.tensor, &repr.indices, &repr.value].into_iter())
            }
            BaseOperationIr::ScatterNd(repr) => {
                Box::new([&repr.data, &repr.indices, &repr.values].into_iter())
            }
            BaseOperationIr::GatherNd(repr) => Box::new([&repr.data, &repr.indices].into_iter()),
            BaseOperationIr::Select(repr) => Box::new([&repr.tensor, &repr.indices].into_iter()),
            BaseOperationIr::SelectAssign(repr) => {
                Box::new([&repr.tensor, &repr.indices, &repr.value].into_iter())
            }
            BaseOperationIr::MaskWhere(repr) => {
                Box::new([&repr.tensor, &repr.mask, &repr.value].into_iter())
            }
            BaseOperationIr::MaskFill(repr) => Box::new([&repr.tensor, &repr.mask].into_iter()),
            BaseOperationIr::Equal(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            BaseOperationIr::EqualElem(repr) => Box::new([&repr.lhs].into_iter()),
            BaseOperationIr::RepeatDim(repr) => Box::new([&repr.tensor].into_iter()),
            BaseOperationIr::Cat(repr) => Box::new(repr.tensors.iter()),
            BaseOperationIr::Cast(repr) => Box::new([&repr.input].into_iter()),
            BaseOperationIr::Unfold(repr) => Box::new([&repr.input].into_iter()),
            BaseOperationIr::Empty(_repr) => Box::new([].into_iter()),
            BaseOperationIr::Ones(_repr) => Box::new([].into_iter()),
            BaseOperationIr::Zeros(_repr) => Box::new([].into_iter()),
            BaseOperationIr::NotEqual(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            BaseOperationIr::NotEqualElem(repr) => Box::new([&repr.lhs].into_iter()),
            BaseOperationIr::All(repr) => Box::new([&repr.input].into_iter()),
            BaseOperationIr::Any(repr) => Box::new([&repr.input].into_iter()),
            BaseOperationIr::AllDim(repr) => Box::new([&repr.input].into_iter()),
            BaseOperationIr::AnyDim(repr) => Box::new([&repr.input].into_iter()),
        }
    }

    fn outputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        match self {
            BaseOperationIr::Reshape(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::SwapDims(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::Permute(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::Expand(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::Flip(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::Slice(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::SliceAssign(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::Gather(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::Scatter(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::ScatterNd(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::GatherNd(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::Select(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::SelectAssign(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::MaskWhere(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::MaskFill(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::Equal(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::EqualElem(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::RepeatDim(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::Cat(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::Cast(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::Unfold(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::Empty(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::Ones(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::Zeros(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::NotEqual(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::NotEqualElem(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::All(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::Any(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::AllDim(repr) => Box::new([&repr.out].into_iter()),
            BaseOperationIr::AnyDim(repr) => Box::new([&repr.out].into_iter()),
        }
    }

    fn mark_read_only(&mut self, nodes: &[TensorId]) -> Vec<TensorIr> {
        let mut output = Vec::new();

        match self {
            BaseOperationIr::Reshape(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::SwapDims(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::Permute(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }

            BaseOperationIr::Expand(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }

            BaseOperationIr::Flip(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::Slice(repr) => {
                repr.tensor.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::SliceAssign(repr) => {
                repr.tensor.mark_read_only(nodes, &mut output);
                repr.value.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::Gather(repr) => {
                repr.tensor.mark_read_only(nodes, &mut output);
                repr.indices.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::Scatter(repr) => {
                repr.tensor.mark_read_only(nodes, &mut output);
                repr.indices.mark_read_only(nodes, &mut output);
                repr.value.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::ScatterNd(repr) => {
                repr.data.mark_read_only(nodes, &mut output);
                repr.indices.mark_read_only(nodes, &mut output);
                repr.values.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::GatherNd(repr) => {
                repr.data.mark_read_only(nodes, &mut output);
                repr.indices.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::Select(repr) => {
                repr.tensor.mark_read_only(nodes, &mut output);
                repr.indices.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::SelectAssign(repr) => {
                repr.tensor.mark_read_only(nodes, &mut output);
                repr.indices.mark_read_only(nodes, &mut output);
                repr.value.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::MaskWhere(repr) => {
                repr.tensor.mark_read_only(nodes, &mut output);
                repr.mask.mark_read_only(nodes, &mut output);
                repr.value.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::MaskFill(repr) => {
                repr.tensor.mark_read_only(nodes, &mut output);
                repr.mask.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::Equal(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::EqualElem(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::RepeatDim(repr) => {
                repr.tensor.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::Cat(repr) => {
                for t in repr.tensors.iter_mut() {
                    t.mark_read_only(nodes, &mut output);
                }
            }
            BaseOperationIr::Cast(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::Unfold(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::Empty(_) => {}
            BaseOperationIr::Zeros(_) => {}
            BaseOperationIr::Ones(_) => {}
            BaseOperationIr::NotEqual(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::NotEqualElem(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::All(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::Any(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::AllDim(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            BaseOperationIr::AnyDim(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
        };

        output
    }

    fn visit_mut(&mut self, v: &mut impl IrVisitorMut) {
        match self {
            BaseOperationIr::Reshape(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::SwapDims(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::Permute(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::Expand(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::Flip(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::Slice(repr) => {
                v.visit_tensor_mut(&mut repr.tensor);
                v.visit_tensor_mut(&mut repr.out);
                repr.ranges.iter_mut().for_each(|r| v.visit_range_mut(r));
            }
            BaseOperationIr::SliceAssign(repr) => {
                v.visit_tensor_mut(&mut repr.tensor);
                v.visit_tensor_mut(&mut repr.value);
                v.visit_tensor_mut(&mut repr.out);
                repr.ranges.iter_mut().for_each(|r| v.visit_range_mut(r));
            }
            BaseOperationIr::Gather(repr) => {
                v.visit_tensor_mut(&mut repr.tensor);
                v.visit_tensor_mut(&mut repr.indices);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::Scatter(repr) => {
                v.visit_tensor_mut(&mut repr.tensor);
                v.visit_tensor_mut(&mut repr.indices);
                v.visit_tensor_mut(&mut repr.value);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::ScatterNd(repr) => {
                v.visit_tensor_mut(&mut repr.data);
                v.visit_tensor_mut(&mut repr.indices);
                v.visit_tensor_mut(&mut repr.values);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::GatherNd(repr) => {
                v.visit_tensor_mut(&mut repr.data);
                v.visit_tensor_mut(&mut repr.indices);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::Select(repr) => {
                v.visit_tensor_mut(&mut repr.tensor);
                v.visit_tensor_mut(&mut repr.indices);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::SelectAssign(repr) => {
                v.visit_tensor_mut(&mut repr.tensor);
                v.visit_tensor_mut(&mut repr.indices);
                v.visit_tensor_mut(&mut repr.value);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::MaskWhere(repr) => {
                v.visit_tensor_mut(&mut repr.tensor);
                v.visit_tensor_mut(&mut repr.mask);
                v.visit_tensor_mut(&mut repr.value);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::MaskFill(repr) => {
                v.visit_tensor_mut(&mut repr.tensor);
                v.visit_tensor_mut(&mut repr.mask);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.value);
            }
            BaseOperationIr::Equal(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::EqualElem(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            BaseOperationIr::RepeatDim(repr) => {
                v.visit_tensor_mut(&mut repr.tensor);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::Cat(repr) => {
                for t in repr.tensors.iter_mut() {
                    v.visit_tensor_mut(t);
                }
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::Cast(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::Unfold(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::Empty(repr) => {
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::Ones(repr) => {
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::Zeros(repr) => {
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::NotEqual(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::NotEqualElem(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            BaseOperationIr::All(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::Any(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::AllDim(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            BaseOperationIr::AnyDim(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
        }
    }
}

impl NumericOperationIr {
    fn inputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        match self {
            NumericOperationIr::Add(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            NumericOperationIr::AddScalar(repr) => Box::new([&repr.lhs].into_iter()),
            NumericOperationIr::Sub(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            NumericOperationIr::SubScalar(repr) => Box::new([&repr.lhs].into_iter()),
            NumericOperationIr::Mul(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            NumericOperationIr::MulScalar(repr) => Box::new([&repr.lhs].into_iter()),
            NumericOperationIr::Div(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            NumericOperationIr::DivScalar(repr) => Box::new([&repr.lhs].into_iter()),
            NumericOperationIr::Rem(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            NumericOperationIr::RemScalar(repr) => Box::new([&repr.lhs].into_iter()),
            NumericOperationIr::GreaterElem(repr) => Box::new([&repr.lhs].into_iter()),
            NumericOperationIr::GreaterEqualElem(repr) => Box::new([&repr.lhs].into_iter()),
            NumericOperationIr::LowerElem(repr) => Box::new([&repr.lhs].into_iter()),
            NumericOperationIr::LowerEqualElem(repr) => Box::new([&repr.lhs].into_iter()),
            NumericOperationIr::Greater(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            NumericOperationIr::GreaterEqual(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            NumericOperationIr::Lower(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            NumericOperationIr::LowerEqual(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            NumericOperationIr::ArgMax(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::ArgTopK(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::TopK(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::ArgMin(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::Clamp(repr) => Box::new([&repr.tensor].into_iter()),
            NumericOperationIr::Abs(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::Full(_repr) => Box::new([].into_iter()),
            NumericOperationIr::MeanDim(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::Mean(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::Sum(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::SumDim(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::Prod(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::ProdDim(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::Max(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::MaxDimWithIndices(repr) => Box::new([&repr.tensor].into_iter()),
            NumericOperationIr::TopKWithIndices(repr) => Box::new([&repr.tensor].into_iter()),
            NumericOperationIr::MinDimWithIndices(repr) => Box::new([&repr.tensor].into_iter()),
            NumericOperationIr::Min(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::MaxDim(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::MinDim(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::MaxAbs(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::MaxAbsDim(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::IntRandom(_repr) => Box::new([].into_iter()),
            NumericOperationIr::Powi(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            NumericOperationIr::PowiScalar(repr) => Box::new([&repr.lhs].into_iter()),
            NumericOperationIr::CumMin(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::CumMax(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::CumProd(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::CumSum(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::Neg(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::Sign(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::ClampMin(repr) => Box::new([&repr.lhs].into_iter()),
            NumericOperationIr::ClampMax(repr) => Box::new([&repr.lhs].into_iter()),
            NumericOperationIr::Sort(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::SortWithIndices(repr) => Box::new([&repr.input].into_iter()),
            NumericOperationIr::ArgSort(repr) => Box::new([&repr.input].into_iter()),
        }
    }

    fn outputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        match self {
            NumericOperationIr::Add(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::AddScalar(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::Sub(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::SubScalar(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::Mul(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::MulScalar(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::Div(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::DivScalar(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::Rem(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::RemScalar(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::GreaterElem(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::GreaterEqualElem(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::LowerElem(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::LowerEqualElem(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::Greater(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::GreaterEqual(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::Lower(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::LowerEqual(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::ArgMax(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::ArgTopK(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::TopK(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::ArgMin(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::Clamp(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::Abs(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::Full(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::MeanDim(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::Mean(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::Sum(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::SumDim(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::Prod(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::ProdDim(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::Max(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::MaxDimWithIndices(repr) => {
                Box::new([&repr.out, &repr.out_indices].into_iter())
            }
            NumericOperationIr::TopKWithIndices(repr) => {
                Box::new([&repr.out, &repr.out_indices].into_iter())
            }
            NumericOperationIr::MinDimWithIndices(repr) => {
                Box::new([&repr.out, &repr.out_indices].into_iter())
            }
            NumericOperationIr::Min(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::MaxDim(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::MinDim(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::MaxAbs(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::MaxAbsDim(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::IntRandom(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::Powi(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::PowiScalar(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::CumMin(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::CumMax(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::CumProd(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::CumSum(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::Neg(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::Sign(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::ClampMin(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::ClampMax(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::Sort(repr) => Box::new([&repr.out].into_iter()),
            NumericOperationIr::SortWithIndices(repr) => {
                Box::new([&repr.out, &repr.out_indices].into_iter())
            }
            NumericOperationIr::ArgSort(repr) => Box::new([&repr.out].into_iter()),
        }
    }
    fn mark_read_only(&mut self, nodes: &[TensorId]) -> Vec<TensorIr> {
        let mut output = Vec::new();

        match self {
            NumericOperationIr::Add(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::AddScalar(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::Sub(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::SubScalar(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::Mul(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::MulScalar(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::Div(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::DivScalar(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::Rem(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::RemScalar(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::GreaterElem(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::GreaterEqualElem(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::LowerElem(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::LowerEqualElem(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::Greater(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::GreaterEqual(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::Lower(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::LowerEqual(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::ArgMax(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::ArgTopK(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::TopK(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::ArgMin(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::Clamp(repr) => {
                repr.tensor.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::Abs(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::Full(_) => {}
            NumericOperationIr::MeanDim(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::Mean(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::Sum(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::SumDim(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::Prod(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::ProdDim(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::Max(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::MaxDimWithIndices(repr) => {
                repr.tensor.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::TopKWithIndices(repr) => {
                repr.tensor.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::MinDimWithIndices(repr) => {
                repr.tensor.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::Min(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::MaxDim(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::MinDim(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::MaxAbs(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::MaxAbsDim(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::IntRandom(_) => {}
            NumericOperationIr::Powi(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::PowiScalar(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::CumSum(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::CumProd(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::CumMin(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::CumMax(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::Neg(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::Sign(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::ClampMin(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::ClampMax(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::Sort(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::SortWithIndices(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            NumericOperationIr::ArgSort(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
        };

        output
    }

    fn visit_mut(&mut self, v: &mut impl IrVisitorMut) {
        match self {
            NumericOperationIr::Add(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::AddScalar(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            NumericOperationIr::Sub(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::SubScalar(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            NumericOperationIr::Mul(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::MulScalar(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            NumericOperationIr::Div(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::DivScalar(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            NumericOperationIr::Rem(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::RemScalar(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            NumericOperationIr::GreaterElem(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            NumericOperationIr::GreaterEqualElem(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            NumericOperationIr::LowerElem(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            NumericOperationIr::LowerEqualElem(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            NumericOperationIr::Greater(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::GreaterEqual(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::Lower(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::LowerEqual(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::ArgMax(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::ArgTopK(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::TopK(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::ArgMin(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::Clamp(repr) => {
                v.visit_tensor_mut(&mut repr.tensor);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.min);
                v.visit_scalar_mut(&mut repr.max);
            }
            NumericOperationIr::Abs(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::Full(repr) => {
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.value);
            }
            NumericOperationIr::MeanDim(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::Mean(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::Sum(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::SumDim(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::Prod(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::ProdDim(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::Max(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::MaxDimWithIndices(repr) => {
                v.visit_tensor_mut(&mut repr.tensor);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_tensor_mut(&mut repr.out_indices);
            }
            NumericOperationIr::TopKWithIndices(repr) => {
                v.visit_tensor_mut(&mut repr.tensor);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_tensor_mut(&mut repr.out_indices);
            }
            NumericOperationIr::MinDimWithIndices(repr) => {
                v.visit_tensor_mut(&mut repr.tensor);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_tensor_mut(&mut repr.out_indices);
            }
            NumericOperationIr::Min(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::MaxDim(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::MinDim(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::MaxAbs(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::MaxAbsDim(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::IntRandom(repr) => {
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::Powi(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::PowiScalar(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            NumericOperationIr::CumMin(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::CumMax(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::CumProd(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::CumSum(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::Neg(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::Sign(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::ClampMin(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            NumericOperationIr::ClampMax(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            NumericOperationIr::Sort(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            NumericOperationIr::SortWithIndices(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_tensor_mut(&mut repr.out_indices);
            }
            NumericOperationIr::ArgSort(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
        }
    }
}

impl FloatOperationIr {
    fn inputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        match self {
            FloatOperationIr::Matmul(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            FloatOperationIr::Cross(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            FloatOperationIr::Random(_repr) => Box::new([].into_iter()),
            FloatOperationIr::Exp(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::Log(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::Log1p(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::Erf(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::Recip(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::PowfScalar(repr) => Box::new([&repr.lhs].into_iter()),
            FloatOperationIr::Sqrt(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::Cos(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::Sin(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::Tanh(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::Round(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::Floor(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::Ceil(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::Trunc(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::IntoInt(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::Quantize(repr) => Box::new(
                [&repr.tensor, &repr.qparams.scales]
                    .into_iter()
                    .chain(repr.qparams.global.iter()),
            ),
            FloatOperationIr::Dequantize(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::IsNan(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::IsInf(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::GridSample2d(repr) => {
                Box::new([&repr.tensor, &repr.grid].into_iter())
            }
            FloatOperationIr::Tan(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::Cosh(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::Sinh(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::ArcCos(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::ArcCosh(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::ArcSin(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::ArcSinh(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::ArcTan(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::ArcTanh(repr) => Box::new([&repr.input].into_iter()),
            FloatOperationIr::ArcTan2(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            FloatOperationIr::Powf(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            FloatOperationIr::Hypot(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
        }
    }
    fn outputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        match self {
            FloatOperationIr::Matmul(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Cross(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Random(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Exp(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Log(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Log1p(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Erf(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Recip(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::PowfScalar(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Sqrt(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Cos(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Sin(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Tanh(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Round(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Floor(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Ceil(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Trunc(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::IntoInt(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Quantize(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Dequantize(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::IsNan(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::IsInf(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::GridSample2d(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Tan(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Cosh(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Sinh(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::ArcCos(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::ArcCosh(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::ArcSin(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::ArcSinh(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::ArcTan(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::ArcTanh(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::ArcTan2(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Powf(repr) => Box::new([&repr.out].into_iter()),
            FloatOperationIr::Hypot(repr) => Box::new([&repr.out].into_iter()),
        }
    }

    fn mark_read_only(&mut self, nodes: &[TensorId]) -> Vec<TensorIr> {
        let mut output = Vec::new();

        match self {
            FloatOperationIr::Matmul(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Cross(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Random(_) => {}
            FloatOperationIr::Exp(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Log(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Log1p(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Erf(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Recip(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::PowfScalar(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Sqrt(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Cos(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Sin(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Tanh(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Round(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Floor(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Ceil(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Trunc(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Quantize(repr) => {
                repr.tensor.mark_read_only(nodes, &mut output);
                repr.qparams.scales.mark_read_only(nodes, &mut output);
                if let Some(global) = &mut repr.qparams.global {
                    global.mark_read_only(nodes, &mut output);
                }
            }
            FloatOperationIr::Dequantize(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::IntoInt(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::IsNan(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::IsInf(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::GridSample2d(repr) => {
                repr.tensor.mark_read_only(nodes, &mut output);
                repr.grid.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Tan(repr) => repr.input.mark_read_only(nodes, &mut output),
            FloatOperationIr::Cosh(repr) => repr.input.mark_read_only(nodes, &mut output),
            FloatOperationIr::Sinh(repr) => repr.input.mark_read_only(nodes, &mut output),
            FloatOperationIr::ArcCos(repr) => repr.input.mark_read_only(nodes, &mut output),
            FloatOperationIr::ArcCosh(repr) => repr.input.mark_read_only(nodes, &mut output),
            FloatOperationIr::ArcSin(repr) => repr.input.mark_read_only(nodes, &mut output),
            FloatOperationIr::ArcSinh(repr) => repr.input.mark_read_only(nodes, &mut output),
            FloatOperationIr::ArcTan(repr) => repr.input.mark_read_only(nodes, &mut output),
            FloatOperationIr::ArcTanh(repr) => repr.input.mark_read_only(nodes, &mut output),
            FloatOperationIr::ArcTan2(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Powf(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            FloatOperationIr::Hypot(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
        };

        output
    }

    fn visit_mut(&mut self, v: &mut impl IrVisitorMut) {
        match self {
            FloatOperationIr::Matmul(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Cross(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Random(repr) => {
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Exp(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Log(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Log1p(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Erf(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Recip(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::PowfScalar(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            FloatOperationIr::Sqrt(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Cos(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Sin(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Tanh(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Round(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Floor(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Ceil(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Trunc(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::IntoInt(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Quantize(repr) => {
                v.visit_tensor_mut(&mut repr.tensor);
                v.visit_tensor_mut(&mut repr.qparams.scales);
                if let Some(global) = &mut repr.qparams.global {
                    v.visit_tensor_mut(global);
                }
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Dequantize(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::IsNan(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::IsInf(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::GridSample2d(repr) => {
                v.visit_tensor_mut(&mut repr.tensor);
                v.visit_tensor_mut(&mut repr.grid);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Tan(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Cosh(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Sinh(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::ArcCos(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::ArcCosh(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::ArcSin(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::ArcSinh(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::ArcTan(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::ArcTanh(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::ArcTan2(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Powf(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            FloatOperationIr::Hypot(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
        }
    }
}

impl IntOperationIr {
    fn inputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        match self {
            IntOperationIr::Matmul(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            IntOperationIr::IntoFloat(repr) => Box::new([&repr.input].into_iter()),
            IntOperationIr::BitwiseAnd(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            IntOperationIr::BitwiseAndScalar(repr) => Box::new([&repr.lhs].into_iter()),
            IntOperationIr::BitwiseOr(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            IntOperationIr::BitwiseOrScalar(repr) => Box::new([&repr.lhs].into_iter()),
            IntOperationIr::BitwiseXor(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            IntOperationIr::BitwiseXorScalar(repr) => Box::new([&repr.lhs].into_iter()),
            IntOperationIr::BitwiseNot(repr) => Box::new([&repr.input].into_iter()),
            IntOperationIr::BitwiseLeftShift(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            IntOperationIr::BitwiseLeftShiftScalar(repr) => Box::new([&repr.lhs].into_iter()),
            IntOperationIr::BitwiseRightShift(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            IntOperationIr::BitwiseRightShiftScalar(repr) => Box::new([&repr.lhs].into_iter()),
        }
    }

    fn outputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        match self {
            IntOperationIr::Matmul(repr) => Box::new([&repr.out].into_iter()),
            IntOperationIr::IntoFloat(repr) => Box::new([&repr.out].into_iter()),
            IntOperationIr::BitwiseAnd(repr) => Box::new([&repr.out].into_iter()),
            IntOperationIr::BitwiseAndScalar(repr) => Box::new([&repr.out].into_iter()),
            IntOperationIr::BitwiseOr(repr) => Box::new([&repr.out].into_iter()),
            IntOperationIr::BitwiseOrScalar(repr) => Box::new([&repr.out].into_iter()),
            IntOperationIr::BitwiseXor(repr) => Box::new([&repr.out].into_iter()),
            IntOperationIr::BitwiseXorScalar(repr) => Box::new([&repr.out].into_iter()),
            IntOperationIr::BitwiseNot(repr) => Box::new([&repr.out].into_iter()),
            IntOperationIr::BitwiseLeftShift(repr) => Box::new([&repr.out].into_iter()),
            IntOperationIr::BitwiseLeftShiftScalar(repr) => Box::new([&repr.out].into_iter()),
            IntOperationIr::BitwiseRightShift(repr) => Box::new([&repr.out].into_iter()),
            IntOperationIr::BitwiseRightShiftScalar(repr) => Box::new([&repr.out].into_iter()),
        }
    }

    fn mark_read_only(&mut self, nodes: &[TensorId]) -> Vec<TensorIr> {
        let mut output = Vec::new();

        match self {
            IntOperationIr::Matmul(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            IntOperationIr::IntoFloat(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            IntOperationIr::BitwiseAnd(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            IntOperationIr::BitwiseAndScalar(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            IntOperationIr::BitwiseOr(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            IntOperationIr::BitwiseOrScalar(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            IntOperationIr::BitwiseXor(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            IntOperationIr::BitwiseXorScalar(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            IntOperationIr::BitwiseNot(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            IntOperationIr::BitwiseLeftShift(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            IntOperationIr::BitwiseLeftShiftScalar(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
            IntOperationIr::BitwiseRightShift(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            IntOperationIr::BitwiseRightShiftScalar(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
            }
        };

        output
    }

    fn visit_mut(&mut self, v: &mut impl IrVisitorMut) {
        match self {
            IntOperationIr::Matmul(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            IntOperationIr::IntoFloat(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            IntOperationIr::BitwiseAnd(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            IntOperationIr::BitwiseAndScalar(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            IntOperationIr::BitwiseOr(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            IntOperationIr::BitwiseOrScalar(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            IntOperationIr::BitwiseXor(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            IntOperationIr::BitwiseXorScalar(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            IntOperationIr::BitwiseNot(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            IntOperationIr::BitwiseLeftShift(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            IntOperationIr::BitwiseLeftShiftScalar(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            IntOperationIr::BitwiseRightShift(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            IntOperationIr::BitwiseRightShiftScalar(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
        }
    }
}

impl BoolOperationIr {
    fn inputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        match self {
            BoolOperationIr::IntoFloat(repr) => Box::new([&repr.input].into_iter()),
            BoolOperationIr::IntoInt(repr) => Box::new([&repr.input].into_iter()),
            BoolOperationIr::Not(repr) => Box::new([&repr.input].into_iter()),
            BoolOperationIr::And(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            BoolOperationIr::Or(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
            BoolOperationIr::Xor(repr) => Box::new([&repr.lhs, &repr.rhs].into_iter()),
        }
    }
    fn outputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        match self {
            BoolOperationIr::IntoFloat(repr) => Box::new([&repr.out].into_iter()),
            BoolOperationIr::IntoInt(repr) => Box::new([&repr.out].into_iter()),
            BoolOperationIr::Not(repr) => Box::new([&repr.out].into_iter()),
            BoolOperationIr::And(repr) => Box::new([&repr.out].into_iter()),
            BoolOperationIr::Or(repr) => Box::new([&repr.out].into_iter()),
            BoolOperationIr::Xor(repr) => Box::new([&repr.out].into_iter()),
        }
    }
    fn mark_read_only(&mut self, nodes: &[TensorId]) -> Vec<TensorIr> {
        let mut output = Vec::new();

        match self {
            BoolOperationIr::IntoFloat(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            BoolOperationIr::IntoInt(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            BoolOperationIr::Not(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
            }
            BoolOperationIr::And(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            BoolOperationIr::Or(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
            BoolOperationIr::Xor(repr) => {
                repr.lhs.mark_read_only(nodes, &mut output);
                repr.rhs.mark_read_only(nodes, &mut output);
            }
        };

        output
    }

    fn visit_mut(&mut self, v: &mut impl IrVisitorMut) {
        match self {
            BoolOperationIr::IntoFloat(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            BoolOperationIr::IntoInt(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            BoolOperationIr::Not(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            BoolOperationIr::And(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            BoolOperationIr::Or(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            BoolOperationIr::Xor(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
        }
    }
}

impl ModuleOperationIr {
    fn inputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        match self {
            ModuleOperationIr::BatchNorm(repr) => {
                Box::new([&repr.x, &repr.gamma, &repr.beta, &repr.mean, &repr.variance].into_iter())
            }
            ModuleOperationIr::Embedding(repr) => {
                Box::new([&repr.weights, &repr.indices].into_iter())
            }
            ModuleOperationIr::EmbeddingBackward(repr) => {
                Box::new([&repr.weights, &repr.out_grad, &repr.indices].into_iter())
            }
            ModuleOperationIr::Linear(repr) => {
                if let Some(bias) = &repr.bias {
                    Box::new([&repr.x, &repr.weight, bias].into_iter())
                } else {
                    Box::new([&repr.x, &repr.weight].into_iter())
                }
            }
            ModuleOperationIr::LinearXBackward(repr) => {
                Box::new([&repr.weight, &repr.output_grad].into_iter())
            }
            ModuleOperationIr::LinearWeightBackward(repr) => {
                Box::new([&repr.x, &repr.output_grad].into_iter())
            }
            ModuleOperationIr::LinearBiasBackward(repr) => {
                Box::new([&repr.output_grad].into_iter())
            }
            ModuleOperationIr::Conv1d(repr) => {
                if let Some(bias) = &repr.bias {
                    Box::new([&repr.x, &repr.weight, bias].into_iter())
                } else {
                    Box::new([&repr.x, &repr.weight].into_iter())
                }
            }
            ModuleOperationIr::Conv1dXBackward(repr) => {
                Box::new([&repr.x, &repr.weight, &repr.output_grad].into_iter())
            }
            ModuleOperationIr::Conv1dWeightBackward(repr) => {
                Box::new([&repr.x, &repr.weight, &repr.output_grad].into_iter())
            }
            ModuleOperationIr::Conv1dBiasBackward(repr) => {
                Box::new([&repr.x, &repr.bias, &repr.output_grad].into_iter())
            }
            ModuleOperationIr::Conv2d(repr) => {
                if let Some(bias) = &repr.bias {
                    Box::new([&repr.x, &repr.weight, bias].into_iter())
                } else {
                    Box::new([&repr.x, &repr.weight].into_iter())
                }
            }
            ModuleOperationIr::Conv2dXBackward(repr) => {
                Box::new([&repr.x, &repr.weight, &repr.output_grad].into_iter())
            }
            ModuleOperationIr::Conv2dWeightBackward(repr) => {
                Box::new([&repr.x, &repr.weight, &repr.output_grad].into_iter())
            }
            ModuleOperationIr::Conv2dBiasBackward(repr) => {
                Box::new([&repr.x, &repr.bias, &repr.output_grad].into_iter())
            }
            ModuleOperationIr::Conv3d(repr) => {
                if let Some(bias) = &repr.bias {
                    Box::new([&repr.x, &repr.weight, bias].into_iter())
                } else {
                    Box::new([&repr.x, &repr.weight].into_iter())
                }
            }
            ModuleOperationIr::Conv3dXBackward(repr) => {
                Box::new([&repr.x, &repr.weight, &repr.output_grad].into_iter())
            }
            ModuleOperationIr::Conv3dWeightBackward(repr) => {
                Box::new([&repr.x, &repr.weight, &repr.output_grad].into_iter())
            }
            ModuleOperationIr::Conv3dBiasBackward(repr) => {
                Box::new([&repr.x, &repr.bias, &repr.output_grad].into_iter())
            }
            ModuleOperationIr::DeformableConv2d(repr) => match (&repr.mask, &repr.bias) {
                (Some(mask), Some(bias)) => {
                    Box::new([&repr.x, &repr.offset, &repr.weight, mask, bias].into_iter())
                }
                (Some(mask), None) => {
                    Box::new([&repr.x, &repr.offset, &repr.weight, mask].into_iter())
                }
                (None, Some(bias)) => {
                    Box::new([&repr.x, &repr.offset, &repr.weight, bias].into_iter())
                }
                (None, None) => Box::new([&repr.x, &repr.offset, &repr.weight].into_iter()),
            },
            ModuleOperationIr::DeformableConv2dBackward(repr) => match (&repr.mask, &repr.bias) {
                (Some(mask), Some(bias)) => Box::new(
                    [
                        &repr.x,
                        &repr.offset,
                        &repr.weight,
                        &repr.out_grad,
                        mask,
                        bias,
                    ]
                    .into_iter(),
                ),
                (Some(mask), None) => Box::new(
                    [&repr.x, &repr.offset, &repr.weight, &repr.out_grad, mask].into_iter(),
                ),
                (None, Some(bias)) => Box::new(
                    [&repr.x, &repr.offset, &repr.weight, &repr.out_grad, bias].into_iter(),
                ),
                (None, None) => {
                    Box::new([&repr.x, &repr.offset, &repr.weight, &repr.out_grad].into_iter())
                }
            },
            ModuleOperationIr::ConvTranspose1d(repr) => {
                if let Some(bias) = &repr.bias {
                    Box::new([&repr.x, &repr.weight, bias].into_iter())
                } else {
                    Box::new([&repr.x, &repr.weight].into_iter())
                }
            }
            ModuleOperationIr::ConvTranspose2d(repr) => {
                if let Some(bias) = &repr.bias {
                    Box::new([&repr.x, &repr.weight, bias].into_iter())
                } else {
                    Box::new([&repr.x, &repr.weight].into_iter())
                }
            }
            ModuleOperationIr::ConvTranspose3d(repr) => {
                if let Some(bias) = &repr.bias {
                    Box::new([&repr.x, &repr.weight, bias].into_iter())
                } else {
                    Box::new([&repr.x, &repr.weight].into_iter())
                }
            }
            ModuleOperationIr::AvgPool1d(repr) => Box::new([&repr.x].into_iter()),
            ModuleOperationIr::AvgPool2d(repr) => Box::new([&repr.x].into_iter()),
            ModuleOperationIr::AvgPool1dBackward(repr) => {
                Box::new([&repr.x, &repr.grad].into_iter())
            }
            ModuleOperationIr::AvgPool2dBackward(repr) => {
                Box::new([&repr.x, &repr.grad].into_iter())
            }
            ModuleOperationIr::AdaptiveAvgPool1d(repr) => Box::new([&repr.x].into_iter()),
            ModuleOperationIr::AdaptiveAvgPool2d(repr) => Box::new([&repr.x].into_iter()),
            ModuleOperationIr::AdaptiveAvgPool1dBackward(repr) => {
                Box::new([&repr.x, &repr.grad].into_iter())
            }
            ModuleOperationIr::AdaptiveAvgPool2dBackward(repr) => {
                Box::new([&repr.x, &repr.grad].into_iter())
            }
            ModuleOperationIr::AdaptiveAvgPool3d(repr) => Box::new([&repr.x].into_iter()),
            ModuleOperationIr::AdaptiveAvgPool3dBackward(repr) => {
                Box::new([&repr.x, &repr.grad].into_iter())
            }
            ModuleOperationIr::MaxPool1d(repr) => Box::new([&repr.x].into_iter()),
            ModuleOperationIr::MaxPool1dWithIndices(repr) => Box::new([&repr.x].into_iter()),
            ModuleOperationIr::MaxPool1dWithIndicesBackward(repr) => {
                Box::new([&repr.x, &repr.indices, &repr.grad].into_iter())
            }
            ModuleOperationIr::MaxPool2d(repr) => Box::new([&repr.x].into_iter()),
            ModuleOperationIr::MaxPool2dWithIndices(repr) => Box::new([&repr.x].into_iter()),
            ModuleOperationIr::MaxPool2dWithIndicesBackward(repr) => {
                Box::new([&repr.x, &repr.indices, &repr.grad].into_iter())
            }
            ModuleOperationIr::Interpolate(repr) => Box::new([&repr.x].into_iter()),
            ModuleOperationIr::InterpolateBackward(repr) => {
                Box::new([&repr.x, &repr.grad].into_iter())
            }
            ModuleOperationIr::Rfft(repr) => Box::new([&repr.signal].into_iter()),
            ModuleOperationIr::IRfft(repr) => {
                Box::new([&repr.input_re, &repr.input_im].into_iter())
            }
            ModuleOperationIr::Attention(repr) => {
                if let Some(mask) = &repr.mask {
                    if let Some(attn_bias) = &repr.attn_bias {
                        Box::new([&repr.query, &repr.key, &repr.value, mask, attn_bias].into_iter())
                    } else {
                        Box::new([&repr.query, &repr.key, &repr.value, mask].into_iter())
                    }
                } else if let Some(attn_bias) = &repr.attn_bias {
                    Box::new([&repr.query, &repr.key, &repr.value, attn_bias].into_iter())
                } else {
                    Box::new([&repr.query, &repr.key, &repr.value].into_iter())
                }
            }
            ModuleOperationIr::CtcLoss(repr) => Box::new(
                [
                    &repr.log_probs,
                    &repr.targets,
                    &repr.input_lengths,
                    &repr.target_lengths,
                ]
                .into_iter(),
            ),
            ModuleOperationIr::CtcLossBackward(repr) => Box::new(
                [
                    &repr.log_probs,
                    &repr.targets,
                    &repr.input_lengths,
                    &repr.target_lengths,
                    &repr.grad_loss,
                ]
                .into_iter(),
            ),
            ModuleOperationIr::LayerNorm(repr) => match &repr.beta {
                Some(beta) => Box::new([&repr.input, &repr.gamma, beta].into_iter()),
                None => Box::new([&repr.input, &repr.gamma].into_iter()),
            },
            ModuleOperationIr::Unfold4d(repr) => Box::new([&repr.x].into_iter()),
            ModuleOperationIr::ConvTranspose1dWeightBackward(repr) => {
                Box::new([&repr.x, &repr.weight, &repr.output_grad].into_iter())
            }
            ModuleOperationIr::ConvTranspose1dBiasBackward(repr) => {
                Box::new([&repr.x, &repr.bias, &repr.output_grad].into_iter())
            }
            ModuleOperationIr::ConvTranspose2dWeightBackward(repr) => {
                Box::new([&repr.x, &repr.weight, &repr.output_grad].into_iter())
            }
            ModuleOperationIr::ConvTranspose2dBiasBackward(repr) => {
                Box::new([&repr.x, &repr.bias, &repr.output_grad].into_iter())
            }
            ModuleOperationIr::ConvTranspose3dWeightBackward(repr) => {
                Box::new([&repr.x, &repr.weight, &repr.output_grad].into_iter())
            }
            ModuleOperationIr::ConvTranspose3dBiasBackward(repr) => {
                Box::new([&repr.x, &repr.bias, &repr.output_grad].into_iter())
            }
        }
    }
    fn outputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        match self {
            ModuleOperationIr::BatchNorm(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::Embedding(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::EmbeddingBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::Linear(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::LinearXBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::LinearWeightBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::LinearBiasBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::Conv1d(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::Conv1dXBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::Conv1dWeightBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::Conv1dBiasBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::Conv2d(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::Conv2dXBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::Conv2dWeightBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::Conv2dBiasBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::Conv3d(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::Conv3dXBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::Conv3dWeightBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::Conv3dBiasBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::DeformableConv2d(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::DeformableConv2dBackward(repr) => {
                match (&repr.mask_grad, &repr.bias_grad) {
                    (Some(mask_grad), Some(bias_grad)) => Box::new(
                        [
                            &repr.input_grad,
                            &repr.offset_grad,
                            &repr.weight_grad,
                            mask_grad,
                            bias_grad,
                        ]
                        .into_iter(),
                    ),
                    (Some(mask_grad), None) => Box::new(
                        [
                            &repr.input_grad,
                            &repr.offset_grad,
                            &repr.weight_grad,
                            mask_grad,
                        ]
                        .into_iter(),
                    ),
                    (None, Some(bias_grad)) => Box::new(
                        [
                            &repr.input_grad,
                            &repr.offset_grad,
                            &repr.weight_grad,
                            bias_grad,
                        ]
                        .into_iter(),
                    ),
                    (None, None) => Box::new(
                        [&repr.input_grad, &repr.offset_grad, &repr.weight_grad].into_iter(),
                    ),
                }
            }
            ModuleOperationIr::ConvTranspose1d(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::ConvTranspose2d(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::ConvTranspose3d(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::AvgPool1d(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::AvgPool2d(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::AvgPool1dBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::AvgPool2dBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::AdaptiveAvgPool1d(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::AdaptiveAvgPool2d(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::AdaptiveAvgPool1dBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::AdaptiveAvgPool2dBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::AdaptiveAvgPool3d(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::AdaptiveAvgPool3dBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::MaxPool1d(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::MaxPool1dWithIndices(repr) => {
                Box::new([&repr.out, &repr.out_indices].into_iter())
            }
            ModuleOperationIr::MaxPool1dWithIndicesBackward(repr) => {
                Box::new([&repr.out].into_iter())
            }
            ModuleOperationIr::MaxPool2d(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::MaxPool2dWithIndices(repr) => {
                Box::new([&repr.out, &repr.out_indices].into_iter())
            }
            ModuleOperationIr::MaxPool2dWithIndicesBackward(repr) => {
                Box::new([&repr.out].into_iter())
            }
            ModuleOperationIr::Interpolate(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::InterpolateBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::Rfft(repr) => Box::new([&repr.out_re, &repr.out_im].into_iter()),
            ModuleOperationIr::IRfft(repr) => Box::new([&repr.out_signal].into_iter()),
            ModuleOperationIr::Attention(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::CtcLoss(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::CtcLossBackward(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::LayerNorm(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::Unfold4d(repr) => Box::new([&repr.out].into_iter()),
            ModuleOperationIr::ConvTranspose1dWeightBackward(repr) => {
                Box::new([&repr.out].into_iter())
            }
            ModuleOperationIr::ConvTranspose1dBiasBackward(repr) => {
                Box::new([&repr.out].into_iter())
            }
            ModuleOperationIr::ConvTranspose2dWeightBackward(repr) => {
                Box::new([&repr.out].into_iter())
            }
            ModuleOperationIr::ConvTranspose2dBiasBackward(repr) => {
                Box::new([&repr.out].into_iter())
            }
            ModuleOperationIr::ConvTranspose3dWeightBackward(repr) => {
                Box::new([&repr.out].into_iter())
            }
            ModuleOperationIr::ConvTranspose3dBiasBackward(repr) => {
                Box::new([&repr.out].into_iter())
            }
        }
    }

    fn mark_read_only(&mut self, nodes: &[TensorId]) -> Vec<TensorIr> {
        let mut output = Vec::new();

        match self {
            ModuleOperationIr::BatchNorm(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.gamma.mark_read_only(nodes, &mut output);
                repr.beta.mark_read_only(nodes, &mut output);
                repr.mean.mark_read_only(nodes, &mut output);
                repr.variance.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::Embedding(repr) => {
                repr.weights.mark_read_only(nodes, &mut output);
                repr.indices.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::EmbeddingBackward(repr) => {
                repr.weights.mark_read_only(nodes, &mut output);
                repr.out_grad.mark_read_only(nodes, &mut output);
                repr.indices.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::Linear(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);

                if let Some(bias) = &mut repr.bias {
                    bias.mark_read_only(nodes, &mut output);
                }
            }
            ModuleOperationIr::LinearXBackward(repr) => {
                repr.weight.mark_read_only(nodes, &mut output);
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::LinearWeightBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::LinearBiasBackward(repr) => {
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::Conv1d(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);

                if let Some(bias) = &mut repr.bias {
                    bias.mark_read_only(nodes, &mut output);
                }
            }
            ModuleOperationIr::Conv1dXBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::Conv1dWeightBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::Conv1dBiasBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.bias.mark_read_only(nodes, &mut output);
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::Conv2d(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);

                if let Some(bias) = &mut repr.bias {
                    bias.mark_read_only(nodes, &mut output);
                }
            }
            ModuleOperationIr::Conv2dXBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::Conv2dWeightBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::Conv2dBiasBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.bias.mark_read_only(nodes, &mut output);
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::Conv3d(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);

                if let Some(bias) = &mut repr.bias {
                    bias.mark_read_only(nodes, &mut output);
                }
            }
            ModuleOperationIr::Conv3dXBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::Conv3dWeightBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::Conv3dBiasBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.bias.mark_read_only(nodes, &mut output);
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::DeformableConv2d(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);
                repr.offset.mark_read_only(nodes, &mut output);

                match (&mut repr.mask, &mut repr.bias) {
                    (Some(mask), Some(bias)) => {
                        mask.mark_read_only(nodes, &mut output);
                        bias.mark_read_only(nodes, &mut output);
                    }
                    (Some(mask), None) => {
                        mask.mark_read_only(nodes, &mut output);
                    }
                    (None, Some(bias)) => {
                        bias.mark_read_only(nodes, &mut output);
                    }
                    (None, None) => {}
                };
            }
            ModuleOperationIr::DeformableConv2dBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);
                repr.offset.mark_read_only(nodes, &mut output);
                repr.out_grad.mark_read_only(nodes, &mut output);

                if let Some(mask) = repr.mask.as_mut() {
                    mask.mark_read_only(nodes, &mut output);
                }
                if let Some(bias) = repr.bias.as_mut() {
                    bias.mark_read_only(nodes, &mut output);
                }
            }
            ModuleOperationIr::ConvTranspose1d(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);

                if let Some(bias) = &mut repr.bias {
                    bias.mark_read_only(nodes, &mut output);
                }
            }
            ModuleOperationIr::ConvTranspose2d(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);

                if let Some(bias) = &mut repr.bias {
                    bias.mark_read_only(nodes, &mut output);
                }
            }
            ModuleOperationIr::ConvTranspose3d(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);

                if let Some(bias) = &mut repr.bias {
                    bias.mark_read_only(nodes, &mut output);
                }
            }
            ModuleOperationIr::AvgPool1d(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::AvgPool2d(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::AvgPool1dBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::AvgPool2dBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::AdaptiveAvgPool1d(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::AdaptiveAvgPool2d(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::AdaptiveAvgPool1dBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::AdaptiveAvgPool2dBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::AdaptiveAvgPool3d(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::AdaptiveAvgPool3dBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::MaxPool1d(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::MaxPool1dWithIndices(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::MaxPool1dWithIndicesBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::MaxPool2d(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::MaxPool2dWithIndices(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::MaxPool2dWithIndicesBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::Interpolate(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::InterpolateBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::Rfft(repr) => {
                repr.signal.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::IRfft(repr) => {
                repr.input_re.mark_read_only(nodes, &mut output);
                repr.input_im.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::Attention(repr) => {
                repr.query.mark_read_only(nodes, &mut output);
                repr.key.mark_read_only(nodes, &mut output);
                repr.value.mark_read_only(nodes, &mut output);
                if let Some(mask) = &mut repr.mask {
                    mask.mark_read_only(nodes, &mut output);
                }
                if let Some(attn_bias) = &mut repr.attn_bias {
                    attn_bias.mark_read_only(nodes, &mut output);
                }
            }
            ModuleOperationIr::CtcLoss(repr) => {
                repr.log_probs.mark_read_only(nodes, &mut output);
                repr.targets.mark_read_only(nodes, &mut output);
                repr.input_lengths.mark_read_only(nodes, &mut output);
                repr.target_lengths.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::CtcLossBackward(repr) => {
                repr.log_probs.mark_read_only(nodes, &mut output);
                repr.targets.mark_read_only(nodes, &mut output);
                repr.input_lengths.mark_read_only(nodes, &mut output);
                repr.target_lengths.mark_read_only(nodes, &mut output);
                repr.grad_loss.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::LayerNorm(repr) => {
                repr.input.mark_read_only(nodes, &mut output);
                repr.gamma.mark_read_only(nodes, &mut output);
                if let Some(beta) = &mut repr.beta {
                    beta.mark_read_only(nodes, &mut output);
                }
            }
            ModuleOperationIr::Unfold4d(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::ConvTranspose1dWeightBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::ConvTranspose1dBiasBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.bias.mark_read_only(nodes, &mut output);
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::ConvTranspose2dWeightBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::ConvTranspose2dBiasBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.bias.mark_read_only(nodes, &mut output);
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::ConvTranspose3dWeightBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.weight.mark_read_only(nodes, &mut output);
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
            ModuleOperationIr::ConvTranspose3dBiasBackward(repr) => {
                repr.x.mark_read_only(nodes, &mut output);
                repr.bias.mark_read_only(nodes, &mut output);
                repr.output_grad.mark_read_only(nodes, &mut output);
            }
        };

        output
    }

    fn visit_mut(&mut self, v: &mut impl IrVisitorMut) {
        match self {
            ModuleOperationIr::BatchNorm(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.gamma);
                v.visit_tensor_mut(&mut repr.beta);
                v.visit_tensor_mut(&mut repr.mean);
                v.visit_tensor_mut(&mut repr.variance);
                v.visit_scalar_mut(&mut repr.epsilon);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::Embedding(repr) => {
                v.visit_tensor_mut(&mut repr.weights);
                v.visit_tensor_mut(&mut repr.indices);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::EmbeddingBackward(repr) => {
                v.visit_tensor_mut(&mut repr.weights);
                v.visit_tensor_mut(&mut repr.out_grad);
                v.visit_tensor_mut(&mut repr.indices);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::Linear(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.weight);
                if let Some(bias) = &mut repr.bias {
                    v.visit_tensor_mut(bias);
                }
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::LinearXBackward(repr) => {
                v.visit_tensor_mut(&mut repr.weight);
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::LinearWeightBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::LinearBiasBackward(repr) => {
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::Conv1d(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.weight);
                if let Some(bias) = &mut repr.bias {
                    v.visit_tensor_mut(bias);
                }
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::Conv1dXBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.weight);
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::Conv1dWeightBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.weight);
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::Conv1dBiasBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.bias);
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::Conv2d(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.weight);
                if let Some(bias) = &mut repr.bias {
                    v.visit_tensor_mut(bias);
                }
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::Conv2dXBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.weight);
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::Conv2dWeightBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.weight);
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::Conv2dBiasBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.bias);
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::Conv3d(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.weight);
                if let Some(bias) = &mut repr.bias {
                    v.visit_tensor_mut(bias);
                }
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::Conv3dXBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.weight);
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::Conv3dWeightBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.weight);
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::Conv3dBiasBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.bias);
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::DeformableConv2d(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.offset);
                v.visit_tensor_mut(&mut repr.weight);
                if let Some(mask) = &mut repr.mask {
                    v.visit_tensor_mut(mask);
                }
                if let Some(bias) = &mut repr.bias {
                    v.visit_tensor_mut(bias);
                }
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::DeformableConv2dBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.offset);
                v.visit_tensor_mut(&mut repr.weight);
                v.visit_tensor_mut(&mut repr.out_grad);
                if let Some(mask) = &mut repr.mask {
                    v.visit_tensor_mut(mask);
                }
                if let Some(bias) = &mut repr.bias {
                    v.visit_tensor_mut(bias);
                }
                v.visit_tensor_mut(&mut repr.input_grad);
                v.visit_tensor_mut(&mut repr.offset_grad);
                v.visit_tensor_mut(&mut repr.weight_grad);
                if let Some(mask_grad) = &mut repr.mask_grad {
                    v.visit_tensor_mut(mask_grad);
                }
                if let Some(bias_grad) = &mut repr.bias_grad {
                    v.visit_tensor_mut(bias_grad);
                }
            }
            ModuleOperationIr::ConvTranspose1d(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.weight);
                if let Some(bias) = &mut repr.bias {
                    v.visit_tensor_mut(bias);
                }
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::ConvTranspose2d(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.weight);
                if let Some(bias) = &mut repr.bias {
                    v.visit_tensor_mut(bias);
                }
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::ConvTranspose3d(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.weight);
                if let Some(bias) = &mut repr.bias {
                    v.visit_tensor_mut(bias);
                }
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::AvgPool1d(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::AvgPool2d(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::AvgPool1dBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::AvgPool2dBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::AdaptiveAvgPool1d(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::AdaptiveAvgPool2d(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::AdaptiveAvgPool1dBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::AdaptiveAvgPool2dBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::AdaptiveAvgPool3d(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::AdaptiveAvgPool3dBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::MaxPool1d(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::MaxPool1dWithIndices(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_tensor_mut(&mut repr.out_indices);
            }
            ModuleOperationIr::MaxPool1dWithIndicesBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.indices);
                v.visit_tensor_mut(&mut repr.grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::MaxPool2d(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::MaxPool2dWithIndices(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_tensor_mut(&mut repr.out_indices);
            }
            ModuleOperationIr::MaxPool2dWithIndicesBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.indices);
                v.visit_tensor_mut(&mut repr.grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::Interpolate(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::InterpolateBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::Rfft(repr) => {
                v.visit_tensor_mut(&mut repr.signal);
                v.visit_tensor_mut(&mut repr.out_re);
                v.visit_tensor_mut(&mut repr.out_im);
            }
            ModuleOperationIr::IRfft(repr) => {
                v.visit_tensor_mut(&mut repr.input_re);
                v.visit_tensor_mut(&mut repr.input_im);
                v.visit_tensor_mut(&mut repr.out_signal);
            }
            ModuleOperationIr::Attention(repr) => {
                v.visit_tensor_mut(&mut repr.query);
                v.visit_tensor_mut(&mut repr.key);
                v.visit_tensor_mut(&mut repr.value);
                if let Some(mask) = &mut repr.mask {
                    v.visit_tensor_mut(mask);
                }
                if let Some(attn_bias) = &mut repr.attn_bias {
                    v.visit_tensor_mut(attn_bias);
                }
                v.visit_tensor_mut(&mut repr.out);
                if let Some(scale) = &mut repr.options.scale {
                    v.visit_scalar_mut(scale);
                }
                if let Some(softcap) = &mut repr.options.softcap {
                    v.visit_scalar_mut(softcap);
                }
            }
            ModuleOperationIr::CtcLoss(repr) => {
                v.visit_tensor_mut(&mut repr.log_probs);
                v.visit_tensor_mut(&mut repr.targets);
                v.visit_tensor_mut(&mut repr.input_lengths);
                v.visit_tensor_mut(&mut repr.target_lengths);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::CtcLossBackward(repr) => {
                v.visit_tensor_mut(&mut repr.log_probs);
                v.visit_tensor_mut(&mut repr.targets);
                v.visit_tensor_mut(&mut repr.input_lengths);
                v.visit_tensor_mut(&mut repr.target_lengths);
                v.visit_tensor_mut(&mut repr.grad_loss);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::LayerNorm(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.gamma);
                if let Some(beta) = &mut repr.beta {
                    v.visit_tensor_mut(beta);
                }
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.epsilon);
            }
            ModuleOperationIr::Unfold4d(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::ConvTranspose1dWeightBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.weight);
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::ConvTranspose1dBiasBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.bias);
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::ConvTranspose2dWeightBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.weight);
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::ConvTranspose2dBiasBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.bias);
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::ConvTranspose3dWeightBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.weight);
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
            ModuleOperationIr::ConvTranspose3dBiasBackward(repr) => {
                v.visit_tensor_mut(&mut repr.x);
                v.visit_tensor_mut(&mut repr.bias);
                v.visit_tensor_mut(&mut repr.output_grad);
                v.visit_tensor_mut(&mut repr.out);
            }
        }
    }
}

impl DistributedOperationIr {
    fn inputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        match self {
            DistributedOperationIr::AllReduce(repr) => Box::new([&repr.tensor].into_iter()),
            DistributedOperationIr::SyncCollective => Box::new([].into_iter()),
        }
    }

    fn outputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        match self {
            DistributedOperationIr::AllReduce(repr) => Box::new([&repr.out].into_iter()),
            DistributedOperationIr::SyncCollective => Box::new([].into_iter()),
        }
    }

    fn mark_read_only(&mut self, nodes: &[TensorId]) -> Vec<TensorIr> {
        let mut output = Vec::new();

        match self {
            DistributedOperationIr::AllReduce(repr) => {
                repr.tensor.mark_read_only(nodes, &mut output);
            }
            DistributedOperationIr::SyncCollective => {}
        }

        output
    }

    fn visit_mut(&mut self, v: &mut impl IrVisitorMut) {
        match self {
            DistributedOperationIr::AllReduce(repr) => {
                v.visit_tensor_mut(&mut repr.tensor);
                v.visit_tensor_mut(&mut repr.out);
            }
            DistributedOperationIr::SyncCollective => {}
        }
    }
}

impl InitOperationIr {
    fn inputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        Box::new([].into_iter())
    }
    fn outputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
        Box::new([&self.out].into_iter())
    }

    fn visit_mut(&mut self, v: &mut impl IrVisitorMut) {
        v.visit_tensor_mut(&mut self.out);
    }
}

impl TensorIr {
    fn mark_read_only(&mut self, nodes: &[TensorId], output: &mut Vec<TensorIr>) {
        if self.status == TensorStatus::ReadWrite && nodes.contains(&self.id) {
            output.push(self.clone());
            self.status = TensorStatus::ReadOnly;
        }
    }
}

impl core::hash::Hash for RandomOpIr {
    fn hash<H: core::hash::Hasher>(&self, state: &mut H) {
        self.out.hash(state);

        match self.distribution {
            Distribution::Default => 1u8.hash(state),
            Distribution::Bernoulli(_) => 2u8.hash(state),
            Distribution::Uniform(_, _) => 3u8.hash(state),
            Distribution::Normal(_, _) => 4u8.hash(state),
        }
    }
}

/// Extension trait to extract outputs when registering an operation.
pub trait OperationOutput<O> {
    /// Extract a single output.
    fn output(self) -> O;

    /// Extract a fixed number of outputs.
    fn outputs<const N: usize>(self) -> [O; N];
}

impl<O: core::fmt::Debug> OperationOutput<O> for Vec<O> {
    fn output(self) -> O {
        let [tensor] = self.outputs();
        tensor
    }

    fn outputs<const N: usize>(self) -> [O; N] {
        self.try_into().unwrap()
    }
}

/// Operation IR for sort along a dim. The output preserves the input shape/dtype.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct SortOpIr {
    /// Input tensor.
    pub input: TensorIr,
    /// Dim along which to sort.
    pub dim: usize,
    /// Sort descending.
    pub descending: bool,
    /// Output tensor (same shape/dtype as input).
    pub out: TensorIr,
}

/// Operation IR for sort-with-indices: returns sorted values + source indices.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct SortWithIndicesOpIr {
    /// Input tensor.
    pub input: TensorIr,
    /// Dim along which to sort.
    pub dim: usize,
    /// Sort descending.
    pub descending: bool,
    /// Output tensor with sorted values.
    pub out: TensorIr,
    /// Output tensor with the indices into the original input.
    pub out_indices: TensorIr,
}

/// Operation IR for layer normalization with optional bias.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct LayerNormOpIr {
    /// Input tensor.
    pub input: TensorIr,
    /// Scale (gamma) parameter.
    pub gamma: TensorIr,
    /// Optional shift (beta) parameter.
    pub beta: Option<TensorIr>,
    /// Numerical-stability epsilon.
    pub epsilon: ScalarIr,
    /// Output tensor.
    pub out: TensorIr,
}

/// Operation IR for unfold4d (a 4d sliding-window kernel-as-explicit-columns reshape).
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct Unfold4dOpIr {
    /// Input tensor of shape `[N, C, H, W]`.
    pub x: TensorIr,
    /// Kernel size `[kH, kW]`.
    pub kernel_size: [usize; 2],
    /// Conv-like options (stride, padding, dilation).
    pub options: Unfold4dOptionsIr,
    /// Output tensor.
    pub out: TensorIr,
}

/// Options for [`Unfold4dOpIr`] — mirrors the backend's `UnfoldOptions`.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct Unfold4dOptionsIr {
    /// Stride `[sH, sW]`.
    pub stride: [usize; 2],
    /// Padding `[pH, pW]`.
    pub padding: [usize; 2],
    /// Dilation `[dH, dW]`.
    pub dilation: [usize; 2],
}

/// Operation IR for the weight-gradient of `conv_transpose1d`.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct ConvTranspose1dWeightBackwardOpIr {
    /// Forward input.
    pub x: TensorIr,
    /// Convolution weights.
    pub weight: TensorIr,
    /// Upstream gradient.
    pub output_grad: TensorIr,
    /// Convolution options.
    pub options: ConvTranspose1dOptionsIr,
    /// Output: gradient w.r.t. `weight`.
    pub out: TensorIr,
}

/// Operation IR for the bias-gradient of `conv_transpose1d`.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct ConvTranspose1dBiasBackwardOpIr {
    /// Forward input.
    pub x: TensorIr,
    /// Bias tensor (sets the output shape).
    pub bias: TensorIr,
    /// Upstream gradient.
    pub output_grad: TensorIr,
    /// Output: gradient w.r.t. `bias`.
    pub out: TensorIr,
}

/// Operation IR for the weight-gradient of `conv_transpose2d`.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct ConvTranspose2dWeightBackwardOpIr {
    /// Forward input.
    pub x: TensorIr,
    /// Convolution weights.
    pub weight: TensorIr,
    /// Upstream gradient.
    pub output_grad: TensorIr,
    /// Convolution options.
    pub options: ConvTranspose2dOptionsIr,
    /// Output: gradient w.r.t. `weight`.
    pub out: TensorIr,
}

/// Operation IR for the bias-gradient of `conv_transpose2d`.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct ConvTranspose2dBiasBackwardOpIr {
    /// Forward input.
    pub x: TensorIr,
    /// Bias tensor.
    pub bias: TensorIr,
    /// Upstream gradient.
    pub output_grad: TensorIr,
    /// Output: gradient w.r.t. `bias`.
    pub out: TensorIr,
}

/// Operation IR for the weight-gradient of `conv_transpose3d`.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct ConvTranspose3dWeightBackwardOpIr {
    /// Forward input.
    pub x: TensorIr,
    /// Convolution weights.
    pub weight: TensorIr,
    /// Upstream gradient.
    pub output_grad: TensorIr,
    /// Convolution options.
    pub options: ConvTranspose3dOptionsIr,
    /// Output: gradient w.r.t. `weight`.
    pub out: TensorIr,
}

/// Operation IR for the bias-gradient of `conv_transpose3d`.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct ConvTranspose3dBiasBackwardOpIr {
    /// Forward input.
    pub x: TensorIr,
    /// Bias tensor.
    pub bias: TensorIr,
    /// Upstream gradient.
    pub output_grad: TensorIr,
    /// Output: gradient w.r.t. `bias`.
    pub out: TensorIr,
}

/// Operation IR for the hard-sigmoid activation function, which takes two scalars
/// (`alpha` and `beta`) in addition to the input tensor.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub struct HardSigmoidOpIr {
    /// Input tensor.
    pub tensor: TensorIr,
    /// Alpha — multiplied with the tensor before adding `beta`.
    pub alpha: ScalarIr,
    /// Beta — added after multiplying by `alpha`.
    pub beta: ScalarIr,
    /// Output tensor.
    pub out: TensorIr,
}

/// Operation intermediate representation for activation functions.
///
/// Activations are kept in their own enum (rather than smeared across `FloatOperationIr`)
/// so backends can match a single arm to dispatch the fused implementation, instead of
/// receiving the function decomposed into 5–20 primitive ops.
#[derive(Clone, Debug, Hash, PartialEq, Serialize, Deserialize)]
pub enum ActivationOperationIr {
    /// [relu](burn_backend::ops::ActivationOps::relu).
    Relu(UnaryOpIr),
    /// [relu_backward](burn_backend::ops::ActivationOps::relu_backward).
    /// `lhs` = output of the forward pass, `rhs` = upstream gradient.
    ReluBackward(BinaryOpIr),
    /// [leaky_relu](burn_backend::ops::ActivationOps::leaky_relu).
    /// `lhs` = input, `rhs` = `negative_slope` scalar.
    LeakyRelu(ScalarOpIr),
    /// [prelu](burn_backend::ops::ActivationOps::prelu).
    /// `lhs` = input, `rhs` = alpha tensor.
    PRelu(BinaryOpIr),
    /// [gelu](burn_backend::ops::ActivationOps::gelu).
    Gelu(UnaryOpIr),
    /// [gelu_backward](burn_backend::ops::ActivationOps::gelu_backward).
    /// `lhs` = forward input, `rhs` = upstream gradient.
    GeluBackward(BinaryOpIr),
    /// [sigmoid](burn_backend::ops::ActivationOps::sigmoid).
    Sigmoid(UnaryOpIr),
    /// [sigmoid_backward](burn_backend::ops::ActivationOps::sigmoid_backward).
    /// `lhs` = output of the forward pass, `rhs` = upstream gradient.
    SigmoidBackward(BinaryOpIr),
    /// [hard_sigmoid](burn_backend::ops::ActivationOps::hard_sigmoid).
    HardSigmoid(HardSigmoidOpIr),
    /// [log_sigmoid](burn_backend::ops::ActivationOps::log_sigmoid).
    LogSigmoid(UnaryOpIr),
    /// [log_sigmoid_backward](burn_backend::ops::ActivationOps::log_sigmoid_backward).
    /// `lhs` = forward input, `rhs` = upstream gradient.
    LogSigmoidBackward(BinaryOpIr),
    /// [softmax](burn_backend::ops::ActivationOps::softmax).
    Softmax(DimOpIr),
    /// [log_softmax](burn_backend::ops::ActivationOps::log_softmax).
    LogSoftmax(DimOpIr),
    /// [softmin](burn_backend::ops::ActivationOps::softmin).
    Softmin(DimOpIr),
}

/// Generate [`ActivationOperationIr`]'s `inputs`/`outputs`/`mark_read_only` from a single
/// variant → input-field table, so the three methods can't drift out of sync. The `out`
/// field is implicit (every variant's op IR has one). Each entry lists the variant and the
/// names of its input tensor fields (`input`, `lhs`/`rhs`, `tensor`, …).
macro_rules! activation_ir_tensor_access {
    ($( $variant:ident => [ $($field:ident),+ ] ),+ $(,)?) => {
        impl ActivationOperationIr {
            fn inputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
                match self {
                    $( Self::$variant(repr) => Box::new([$(&repr.$field),+].into_iter()), )+
                }
            }

            fn outputs(&self) -> Box<dyn Iterator<Item = &TensorIr> + '_> {
                match self {
                    $( Self::$variant(repr) => Box::new([&repr.out].into_iter()), )+
                }
            }

            fn mark_read_only(&mut self, nodes: &[TensorId]) -> Vec<TensorIr> {
                let mut output = Vec::new();
                match self {
                    $( Self::$variant(repr) => {
                        $( repr.$field.mark_read_only(nodes, &mut output); )+
                    } )+
                }
                output
            }
        }
    };
}

impl ActivationOperationIr {
    fn visit_mut(&mut self, v: &mut impl IrVisitorMut) {
        match self {
            ActivationOperationIr::Relu(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            ActivationOperationIr::ReluBackward(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            ActivationOperationIr::LeakyRelu(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.rhs);
            }
            ActivationOperationIr::PRelu(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            ActivationOperationIr::Gelu(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            ActivationOperationIr::GeluBackward(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            ActivationOperationIr::Sigmoid(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            ActivationOperationIr::SigmoidBackward(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            ActivationOperationIr::HardSigmoid(repr) => {
                v.visit_tensor_mut(&mut repr.tensor);
                v.visit_tensor_mut(&mut repr.out);
                v.visit_scalar_mut(&mut repr.alpha);
                v.visit_scalar_mut(&mut repr.beta);
            }
            ActivationOperationIr::LogSigmoid(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            ActivationOperationIr::LogSigmoidBackward(repr) => {
                v.visit_tensor_mut(&mut repr.lhs);
                v.visit_tensor_mut(&mut repr.rhs);
                v.visit_tensor_mut(&mut repr.out);
            }
            ActivationOperationIr::Softmax(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            ActivationOperationIr::LogSoftmax(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
            ActivationOperationIr::Softmin(repr) => {
                v.visit_tensor_mut(&mut repr.input);
                v.visit_tensor_mut(&mut repr.out);
            }
        }
    }
}

activation_ir_tensor_access! {
    Relu => [input],
    ReluBackward => [lhs, rhs],
    LeakyRelu => [lhs],
    PRelu => [lhs, rhs],
    Gelu => [input],
    GeluBackward => [lhs, rhs],
    Sigmoid => [input],
    SigmoidBackward => [lhs, rhs],
    HardSigmoid => [tensor],
    LogSigmoid => [input],
    LogSigmoidBackward => [lhs, rhs],
    Softmax => [input],
    LogSoftmax => [input],
    Softmin => [input],
}

#[cfg(test)]
mod visit_mut_tests {
    use super::*;
    use burn_backend::{DType, Shape};

    fn tensor(id: u64) -> TensorIr {
        TensorIr::uninit(TensorId::new(id), Shape::from([2, 2]), DType::F32)
    }

    /// Bumps every visited tensor id by 100 and collects (and rewrites) every visited scalar.
    #[derive(Default)]
    struct CollectVisitor {
        scalars: Vec<ScalarIr>,
        rewrite_scalar: Option<ScalarIr>,
    }

    impl IrVisitorMut for CollectVisitor {
        fn visit_tensor_mut(&mut self, tensor: &mut TensorIr) {
            tensor.id = TensorId::new(tensor.id.value() + 100);
        }

        fn visit_scalar_mut(&mut self, scalar: &mut ScalarIr) {
            self.scalars.push(*scalar);
            if let Some(value) = self.rewrite_scalar {
                *scalar = value;
            }
        }
    }

    #[test]
    fn visit_mut_visits_all_tensors_and_scalars() {
        // NumericFloat MulScalar (a ScalarOpIr with a scalar `rhs`).
        let mut mul = OperationIr::NumericFloat(
            DType::F32,
            NumericOperationIr::MulScalar(ScalarOpIr {
                lhs: tensor(1),
                rhs: ScalarIr::Float(2.0),
                out: tensor(2),
            }),
        );

        // Bump every tensor id by 100 and rewrite the scalar to 9.0.
        let mut visitor = CollectVisitor {
            rewrite_scalar: Some(ScalarIr::Float(9.0)),
            ..Default::default()
        };
        mul.visit_mut(&mut visitor);

        let ids: Vec<u64> = mul
            .inputs()
            .chain(mul.outputs())
            .map(|t| t.id.value())
            .collect();
        assert_eq!(ids, vec![101, 102]);
        // The scalar must have been visited (and was rewritable).
        assert_eq!(visitor.scalars, vec![ScalarIr::Float(2.0)]);

        // Visiting again observes the rewritten scalar.
        let mut after = CollectVisitor::default();
        mul.visit_mut(&mut after);
        assert_eq!(after.scalars, vec![ScalarIr::Float(9.0)]);

        // BaseFloat Reshape (a ShapeOpIr, input + out, no scalars).
        let mut reshape = OperationIr::BaseFloat(BaseOperationIr::Reshape(ShapeOpIr {
            input: tensor(10),
            out: tensor(11),
        }));

        let mut visitor = CollectVisitor::default();
        reshape.visit_mut(&mut visitor);
        let ids: Vec<u64> = reshape
            .inputs()
            .chain(reshape.outputs())
            .map(|t| t.id.value())
            .collect();
        assert_eq!(ids, vec![110, 111]);
        assert_eq!(visitor.scalars.len(), 0);
    }
}