// binary zip operation, e.g. Plus
// If allowBroadcast then one can be a sub-dimension of the other (if layout then only for rows, otherwise for cols, too).
// This also helpfully resizes the children if not yet sized.
void ComputationNodeBase::ValidateBinaryZip(bool isFinalValidationPass, bool allowBroadcast)
{
assert(m_inputs.size() == 2);
ComputationNodeBase::Validate(isFinalValidationPass);
InferMBLayoutFromInputsForStandardCase(isFinalValidationPass);
ValidateInferBinaryInputDims();
if (isFinalValidationPass)
ValidateMBLayout(Input(0), Input(1));
// result has tensor shape with dimensions being the max over both
let shape0 = GetInputSampleLayout(0);
let shape1 = GetInputSampleLayout(1);
SmallVector<size_t> dims = shape0.GetDims();
if (shape1.GetRank() > dims.size())
dims.resize(shape1.GetRank(), 1); // pad with ones
// If rank of [0] is higher than we only need to take max over rank [1].
// If rank of [1] is higher then we have padded to equal lentgh.
for (size_t k = 0; k < shape1.GetRank(); k++)
{
size_t dim1 = shape1[k];
// BUGBUG: We must consider the allowBroadcast flag here.
if (dims[k] <= 1 && dim1 != 0) // is [0] broadcasting (1) or unspecified (0)?
dims[k] = dim1; // then use dimension we broadcast to
else if (dim1 <= 1 && dims[k] != 0) // if [1] is broadcasting or unspecified
; // then dims is already correct
else if (isFinalValidationPass && dim1 != dims[k]) // no broadcasting or unspecified: they must match
InvalidArgument("%ls: Input dimensions [%s] and [%s] are not compatible.",
NodeDescription().c_str(), string(shape0).c_str(), string(shape1).c_str());
}
SetDims(TensorShape(dims), HasMBLayout());
}
/*virtual*/ void ReduceElementsNode<ElemType>::Validate(bool isFinalValidationPass) /*override*/
{
Base::Validate(isFinalValidationPass);
InferMBLayoutFromInputsForStandardCase(isFinalValidationPass);
// validate the opcode (in case we got instantiated empty and never updated)
ValidateOp();
let shape = Input(0)->GetSampleLayout();
auto dims = shape.GetDims();
size_t reducedDim = 0; // (init to keep compiler happy)
if (m_axis == 0)
{
reducedDim = shape.GetNumElements();
dims = { 1 }; // entire sample is reduced to a scalar
}
else if (m_axis - 1 >= 0 && m_axis - 1 < dims.size())
{
reducedDim = dims[m_axis - 1];
dims[m_axis - 1] = 1; // one axis is reduced to a scalar
}
else if (isFinalValidationPass)
InvalidArgument("The shape of %ls [%s] has no axis %d", NodeDescription().c_str(), string(shape).c_str(), m_axis);
// for "Mean", we must divide by #elements
if (isFinalValidationPass && m_operation == L"Mean")
m_scale = (ElemType)(1.0 / reducedDim);
else
m_scale = (ElemType)1;
SetDims(TensorShape(dims), Input(0)->HasMBLayout());
}
// binary reduce-to-(1,1) operation, e.g. CrossEntropyWithSoftmaxNode
// Currently only called by criterion nodes.
// This function also infers child LearnableParameters. In case you wonder why this is needed for criterion nodes, there are edge cases, e.g. a
// learnable parameter being regularized by a criterion node, where the learnable parameter is fed both into that criterion node and other places.
void ComputationNodeBase::ValidateBinaryReduce(bool isFinalValidationPass)
{
ComputationNodeBase::Validate(isFinalValidationPass);
m_pMBLayout = nullptr; // this node does not hold mini-batch data
ValidateInferBinaryInputDims();
if (isFinalValidationPass)
{
if (!(Input(0)->GetSampleLayout().IsElementwiseCompatibleWith(Input(1)->GetSampleLayout())))
{
string s1 = Input(0)->GetSampleLayout();
string s2 = Input(1)->GetSampleLayout();
// BUGBUG: Allow broadcasting?
LogicError("%ls: The tensor dimensions in the inputs do not match. %s != %s", NodeDescription().c_str(), s1.c_str(), s2.c_str());
}
else if (!(Input(0)->HasMBLayout()))
LogicError("%ls: Expected MBLayout in Input 0.", NodeDescription().c_str());
else if (!(Input(1)->HasMBLayout()))
LogicError("%ls: Expected MBLayout in Input 1.", NodeDescription().c_str());
// Shape of the MBLayouts is checked at runtime.
}
SetDims(TensorShape(1), false);
}
/*virtual*/ void ReduceElementsNode<ElemType>::Validate(bool isFinalValidationPass) /*override*/
{
Base::Validate(isFinalValidationPass);
InferMBLayoutFromInputsForStandardCase(isFinalValidationPass);
// validate the opcode (in case we got instantiated empty and never updated)
ValidateOp();
let shape = Input(0)->GetSampleLayout();
auto dims = shape.GetDims();
if (m_axis == 0)
dims = { 1 }; // entire sample is reduced to a scalar
else if (m_axis - 1 >= 0 && m_axis - 1 < dims.size())
dims[m_axis - 1] = 1; // one axis is reduced to a scalar
else if (isFinalValidationPass)
InvalidArgument("The shape of %ls [%s] has no axis %d", NodeDescription().c_str(), string(shape).c_str(), m_axis);
SetDims(TensorShape(dims), Input(0)->HasMBLayout());
}
static bool AddNodeValue (TRI_json_t* row, TRI_aql_node_t* const node) {
TRI_json_t* result;
TRI_json_t* type;
TRI_json_t* value;
result = TRI_CreateArrayJson(TRI_UNKNOWN_MEM_ZONE);
if (result == NULL) {
return false;
}
type = NodeType(node);
if (type != NULL) {
TRI_Insert3ArrayJson(TRI_UNKNOWN_MEM_ZONE,
result,
"type",
type);
}
value = NodeDescription(node);
if (value != NULL) {
TRI_Insert3ArrayJson(TRI_UNKNOWN_MEM_ZONE,
result,
"value",
value);
}
if (node->_type == TRI_AQL_NODE_COLLECTION) {
TRI_json_t* extra = TRI_GetJsonCollectionHintAql(TRI_AQL_NODE_DATA(node));
if (extra != NULL) {
TRI_Insert3ArrayJson(TRI_UNKNOWN_MEM_ZONE,
result,
"extra",
extra);
}
}
TRI_Insert3ArrayJson(TRI_UNKNOWN_MEM_ZONE, row, "expression", result);
return true;
}
// binary zip operation, e.g. Plus
// If allowBroadcast then one can be a sub-dimension of the other (if layout then only for rows, otherwise for cols, too).
// This also helpfully resizes the children if not yet sized.
void ComputationNodeBase::ValidateBinaryZip(bool isFinalValidationPass, bool allowBroadcast)
{
assert(m_inputs.size() == 2);
ComputationNodeBase::Validate(isFinalValidationPass);
InferMBLayoutFromInputsForStandardCase(isFinalValidationPass);
ValidateInferBinaryInputDims();
if (isFinalValidationPass &&
Input(0)->GetMBLayout() != Input(1)->GetMBLayout() && Input(0)->HasMBLayout() && Input(1)->HasMBLayout())
{
LogicError("%ls: Minibatch layouts are not the same between arguments and might get out of sync during runtime. If this is by design, use ReconcileDynamicAxis() to forward layouts between nodes.", NodeDescription().c_str());
}
// result has tensor shape with dimensions being the max over both
let shape0 = GetInputSampleLayout(0);
let shape1 = GetInputSampleLayout(1);
SmallVector<size_t> dims = shape0.GetDims();
if (shape1.GetRank() > dims.size())
dims.resize(shape1.GetRank(), 1); // pad with ones
// If rank of [0] is higher than we only need to take max over rank [1].
// If rank of [1] is higher then we have padded to equal lentgh.
for (size_t k = 0; k < shape1.GetRank(); k++)
{
size_t dim1 = shape1[k];
// BUGBUG: We must consider the allowBroadcast flag here.
if (dims[k] == 1) // is [0] broadcasting?
dims[k] = dim1; // then use dimension we broadcast to
else if (dim1 == 1) // if [1] is broadcasting
; // dims is already correct
else if (isFinalValidationPass && dim1 != dims[k]) // no broadcasting: they must match
InvalidArgument("%ls: Input dimensions [%s] and [%s] are not compatible.",
NodeDescription().c_str(), string(shape0).c_str(), string(shape1).c_str());
}
SetDims(TensorShape(dims), HasMBLayout());
}
void DescriptionsManager::processMessage(const Message* message)
{
// if we have a disconnection message
{
const Disconnected *disconnected = dynamic_cast<const Disconnected *>(message);
if (disconnected)
{
NodesDescriptionsMap::iterator it = nodesDescriptions.find(disconnected->source);
if (it != nodesDescriptions.end())
nodesDescriptions.erase(it);
}
}
// if we have an initial description
{
const Description *description = dynamic_cast<const Description *>(message);
if (description)
{
NodesDescriptionsMap::iterator it = nodesDescriptions.find(description->source);
// We can receive a description twice, for instance if there is another IDE connected
if (it != nodesDescriptions.end())
return;
// Call a user function when a node protocol version mismatches
if (description->protocolVersion != ASEBA_PROTOCOL_VERSION)
{
nodeProtocolVersionMismatch(description->name, description->protocolVersion);
return;
}
// create node and copy description into it
nodesDescriptions[description->source] = NodeDescription(*description);
checkIfNodeDescriptionComplete(description->source, nodesDescriptions[description->source]);
}
}
// if we have a named variabledescription
{
const NamedVariableDescription *description = dynamic_cast<const NamedVariableDescription *>(message);
if (description)
{
NodesDescriptionsMap::iterator it = nodesDescriptions.find(description->source);
// we must have received a description first
if (it == nodesDescriptions.end())
return;
// copy description into array if array is empty
if (it->second.namedVariablesReceptionCounter < it->second.namedVariables.size())
{
it->second.namedVariables[it->second.namedVariablesReceptionCounter++] = *description;
checkIfNodeDescriptionComplete(it->first, it->second);
}
}
}
// if we have a local event description
{
const LocalEventDescription *description = dynamic_cast<const LocalEventDescription *>(message);
if (description)
{
NodesDescriptionsMap::iterator it = nodesDescriptions.find(description->source);
// we must have received a description first
if (it == nodesDescriptions.end())
return;
// copy description into array if array is empty
if (it->second.localEventsReceptionCounter < it->second.localEvents.size())
{
it->second.localEvents[it->second.localEventsReceptionCounter++] = *description;
checkIfNodeDescriptionComplete(it->first, it->second);
}
}
}
// if we have a native function description
{
const NativeFunctionDescription *description = dynamic_cast<const NativeFunctionDescription *>(message);
if (description)
{
NodesDescriptionsMap::iterator it = nodesDescriptions.find(description->source);
// we must have received a description first
if (it == nodesDescriptions.end())
return;
// copy description into array
if (it->second.nativeFunctionReceptionCounter < it->second.nativeFunctions.size())
{
it->second.nativeFunctions[it->second.nativeFunctionReceptionCounter++] = *description;
checkIfNodeDescriptionComplete(it->first, it->second);
}
}
}
}
/*virtual*/ void GatherPackedNode<ElemType>::Validate(bool isFinalValidationPass) /*override*/
{
ComputationNodeBase::Validate(isFinalValidationPass);
// inherit MBLayout from indexData
m_pMBLayout = Input(INDEXDATA)->GetMBLayout();
if (isFinalValidationPass && (!Input(INDEXDATA)->HasMBLayout()))
LogicError("%ls requires first argument (index data) to have a time dimension.", NodeDescription().c_str());
bool sourceHasTimeDimension = Input(SOURCEDATA)->HasMBLayout();
if (isFinalValidationPass && Input(INDEXDATA)->GetSampleLayout().GetNumElements() != 1)
InvalidArgument("%ls requires the first argument (index data) to be a scalar time sequence.", NodeDescription().c_str());
// inherit tensor dimension from sourceData, minus the last (column or time) dimension. TODO this needs to become simpler...
if (sourceHasTimeDimension)
SetDims(Input(SOURCEDATA)->GetSampleLayout(), HasMBLayout());
else
{
SmallVector<size_t> layout = { 1 }; // Scalar
if (Input(SOURCEDATA)->GetSampleLayout().GetRank() > 1)
{
auto srcLayout = Input(SOURCEDATA)->GetSampleLayout().GetDims();
layout.assign(srcLayout.begin(), srcLayout.end() - 1);
}
SetDims(TensorShape(layout), HasMBLayout());
}
}
void ReduceElementsNode<ElemType>::ValidateOp()
{
#if 1 // legacy with initial experiments, delete this soon
if (m_operation == L"Plus") m_reductionOp = ElementWiseOperator::opSum;
else
#endif
if (m_operation == L"Sum" ) m_reductionOp = ElementWiseOperator::opSum;
else if (m_operation == L"Max") m_reductionOp = ElementWiseOperator::opMax;
else if (m_operation == L"Min") m_reductionOp = ElementWiseOperator::opMin;
// more here
else InvalidArgument("%ls was given an invalid operation code '%ls'. Allowed are: 'Sum', 'Max', 'Min'.", NodeDescription().c_str(), m_operation.c_str());
}
// N-nary zip operation, e.g. for TernaryZip for clip()
// If allowBroadcast then one can be a sub-dimension of the other (if layout then only for rows, otherwise for cols, too).
// This also helpfully resizes the children if not yet sized.
void ComputationNodeBase::ValidateNaryZip(bool isFinalValidationPass, bool allowBroadcast, size_t numInputs)
{
assert(m_inputs.size() == numInputs);
ComputationNodeBase::Validate(isFinalValidationPass);
InferMBLayoutFromInputsForStandardCase(isFinalValidationPass);
ValidateInferNaryInputDims(numInputs);
// check minibatch layout consistency for all possible pairs (n choose 2)
if (isFinalValidationPass)
for (size_t i = 0; i < numInputs; i++)
for (size_t j = i + 1; j < numInputs; j++)
ValidateMBLayout(Input(i), Input(j));
// result has tensor shape with dimensions being the max over all inputs
let shape0 = GetInputSampleLayout(0);
// dims is max over all inputs
size_t maxRank = shape0.GetRank();
for (size_t i = 1; i < numInputs; i++)
{
let shape = GetInputSampleLayout(i);
if (shape.GetRank() > maxRank)
maxRank = shape.GetRank();
}
SmallVector<size_t> dims = shape0.GetDims();
dims.resize(maxRank, 1); // pad with 1
// first check for invalid dimensions
for (size_t k = 0; k < maxRank; k++)
{
size_t maxDim = 0;
TensorShape maxShape = shape0; // arbitrary; this is just used for the error message
for (size_t i = 0; i < numInputs; i++)
{
let currentShape = GetInputSampleLayout(i);
size_t currentRank = currentShape.GetRank();
// make sure that the rank of this input is bigger than the current index (otherwise, these are implied singleton dimensions that do not need to be checked)
if (currentRank > k)
{
size_t currentDim = currentShape[k];
if (currentDim > 1 && maxDim != currentDim && maxDim > 1) // 1=broadcasting, 0=not known yet, meant to be inferred
{
InvalidArgument("%ls: Input dimensions [%s] and [%s] are not compatible.",
NodeDescription().c_str(), string(maxShape).c_str(), string(currentShape).c_str());
}
else if (currentDim > maxDim)
{
maxDim = currentDim;
maxShape = currentShape;
}
}
}
}
// now set up the right dims
for (size_t k = 0; k < maxRank; k++)
{
for (size_t i = 0; i < numInputs; i++)
{
let shape = GetInputSampleLayout(i);
if (shape.GetRank() > k)
{
size_t dim = shape[k];
if (dims[k] <= 1 && dim != 0)
dims[k] = dim;
}
}
}
SetDims(TensorShape(dims), HasMBLayout());
}
void ReduceElementsNode<ElemType>::ValidateOp()
{
if (m_operation == L"Plus") m_op = ElementWiseOperator::opSum;
// more here
else InvalidArgument("%ls was given an invalid operation code '%ls'. Allowed are: 'Plus'. And a few more soon.", NodeDescription().c_str(), m_operation.c_str());
}
