Class Node

• Defined in File node.h

Inheritance Relationships

Base Type

• public marian::Chainable< Tensor > (Template Class Chainable)

Derived Types

• public marian::ConstantNode (Struct ConstantNode)

• public marian::NaryNodeOp (Struct NaryNodeOp)

• public marian::ParamNode (Struct ParamNode)

Class Documentation

class Node : public marian::Chainable<Tensor>

Main node class for computation graph, implements most common functions demanded by Chainable.

Each operation in a computation graph is a node.

Subclassed by marian::ConstantNode, marian::NaryNodeOp, marian::ParamNode

Public Functions

Node(Ptr<ExpressionGraph> graph, const Shape &shape, const Type &valueType = Type::float32)

virtual ~Node()

float scalar()

virtual NodeOps forwardOps()

virtual NodeOps backwardOps()

virtual void runForward(const NodeOps &ops)

virtual void runBackward(const NodeOps &ops)

void forward()

void backward()

virtual bool trainable()

virtual void setTrainable(bool trainable)

virtual bool memoize()

virtual void setMemoize(bool memoize)

virtual void setId(size_t id)

virtual size_t getId()

virtual void increaseEdges(size_t edges = 1)

virtual void decreaseEdges(size_t edges = 1)

virtual size_t edges()

virtual Ptr<ExpressionGraph> graph()

virtual void debug(const std::string &message)

virtual bool marked_for_debug()

virtual const std::string &debug_message()

void allocate()

void free()

virtual void init()

void init_dependent()

Initialization for backward step of top node in computation graph.

Allocates memory and sets gradient to 1 (df/df == 1).

void set_zero_adjoint()

Initialization for backward step of any non-top node in computation graph.

Allocates memory and sets gradient to 0 for further accumulation of gradients from all parents.

virtual Tensor &val()

virtual Tensor &grad()

virtual const Shape &shape()

virtual const Type &value_type()

void set_name(const std::string &name)

const std::string &name() const

virtual const std::string form()

virtual const std::string color()

virtual const std::string label()

virtual std::string graphviz()

virtual std::vector<Expr> &children()

virtual Expr child(size_t i)

Ptr<Backend> getBackend()

void record(Ptr<AutoTunerRecorder> recorder, size_t recorderHash, bool stop)

virtual void markCheckpoint()

virtual bool isCheckpoint() const

virtual void setSubtape(Ptr<std::list<Expr>> subtape)

virtual Ptr<std::list<Expr>> getSubtape()

Protected Attributes

size_t id_ = {0}

size_t edges_ = {0}

bool trainable_ = {true}

bool destroy_ = {true}

bool memoize_ = {false}

std::vector<Expr> children_

Weak<ExpressionGraph> graph_

Shape shape_ = {1, 1, 1, 1}

Type valueType_ = {Type::float32}

std::string name_ = {"none"}

Tensor val_ = {nullptr}

Tensor adj_ = {nullptr}

bool markedForDebug_ = {false}

std::string debugMessage_

Ptr<std::list<Expr>> subtape_

bool isCheckpoint_ = {false}

Ptr<AutoTunerRecorder> recorder_

size_t recorderHash_

bool recorderStop_