Enum OperationIr 

pub enum OperationIr {
    BaseFloat(BaseOperationIr),
    BaseInt(BaseOperationIr),
    BaseBool(BaseOperationIr),
    NumericFloat(DType, NumericOperationIr),
    NumericInt(DType, NumericOperationIr),
    Bool(BoolOperationIr),
    Int(IntOperationIr),
    Float(DType, FloatOperationIr),
    Module(ModuleOperationIr),
    Init(InitOperationIr),
    Custom(CustomOpIr),
    Drop(TensorIr),
    Distributed(DistributedOperationIr),
    Activation(ActivationOperationIr),
}

Describe all tensor operations possible.

Variants

BaseFloat(BaseOperationIr)

Basic operation on a float tensor.

BaseInt(BaseOperationIr)

Basic operation on an int tensor.

BaseBool(BaseOperationIr)

Basic operation on a bool tensor.

NumericFloat(DType, NumericOperationIr)

Numeric operation on a float tensor.

NumericInt(DType, NumericOperationIr)

Numeric operation on an int tensor.

Bool(BoolOperationIr)

Operation specific to a bool tensor.

Int(IntOperationIr)

Operation specific to an int tensor.

Float(DType, FloatOperationIr)

Operation specific to a float tensor.

Module(ModuleOperationIr)

Module operation.

Init(InitOperationIr)

Initialize operation.

Custom(CustomOpIr)

A custom operation.

Drop(TensorIr)

A tensor is dropped.

Distributed(DistributedOperationIr)

Operation specific to a distributed tensor.

Activation(ActivationOperationIr)

Activation function operation (relu, gelu, sigmoid, softmax, …).

Implementations

impl OperationIr

pub fn inputs(&self) -> impl Iterator<Item = &TensorIr>

Get all input tensors involved with the current operation.

pub fn outputs(&self) -> impl Iterator<Item = &TensorIr>

Get all output tensors involved with the current operation.

pub fn nodes(&self) -> Vec<&TensorIr>

Get all tensor involved with the current operation.

pub fn mark_read_only(&mut self, nodes: &[TensorId]) -> Vec<TensorIr>

Set the given nodes that are read write to read only in the current operation.

Returns the tensor that were updated with their original representation.

pub fn visit_mut(&mut self, v: &mut impl IrVisitorMut)

Visit every mutable component of this operation in place.

For each variant, visits its tensors (in the same order as inputs then outputs combined), then its scalars, then its slice ranges. Only Slice/SliceAssign (under the Base* variants) carry ranges; every other category visits no ranges.

Trait Implementations

impl Clone for OperationIr

fn clone(&self) -> OperationIr

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for OperationIr

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for OperationIr

fn deserialize<__D>( __deserializer: __D, ) -> Result<OperationIr, <__D as Deserializer<'de>>::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Hash for OperationIr

fn hash<__H>(&self, state: &mut __H)

where __H: Hasher,

Feeds this value into the given Hasher.

fn hash_slice<H>(data: &[Self], state: &mut H)

where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher.

impl PartialEq for OperationIr

fn eq(&self, other: &OperationIr) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Serialize for OperationIr

fn serialize<__S>( &self, __serializer: __S, ) -> Result<<__S as Serializer>::Ok, <__S as Serializer>::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for OperationIr