提交 925aceac 编写于 作者: M mindspore-ci-bot 提交者: Gitee

!4101 fix op multi output problem

Merge pull request !4101 from yeyunpeng2020/master_split
......@@ -189,7 +189,9 @@ union PrimitiveType {
ActivationGrad,
PriorBox,
SpaceToBatchND,
TopKV2
TopKV2,
Return,
MakeTuple
}
enum QuantType: int {
......
......@@ -864,3 +864,9 @@ table TopKV2 {
sorted : bool = true;
}
table MakeTuple {
}
table Return {
}
\ No newline at end of file
......@@ -81,8 +81,7 @@ bool AnfExporter::RemoveIfTupleGetItem(const CNodePtr &cnode) {
return false;
}
ValueNodePtr valueNode = utils::cast<ValueNodePtr>(indexNode);
mapRemoveGetItem_[tupleGetItemNode->input(1)->fullname_with_scope()] =
GetValue<int>(valueNode->value());
mapRemoveGetItem_[tupleGetItemNode->input(1)->fullname_with_scope()] = GetValue<int>(valueNode->value());
} else {
inputs.emplace_back(cnode->input(i));
}
......@@ -114,16 +113,34 @@ schema::MetaGraphT *AnfExporter::Export(const FuncGraphPtr &funcGraph) {
auto metaGraphT = std::make_unique<schema::MetaGraphT>();
for (const auto &cnode : cnodes) {
auto primitive = GetValueNode<PrimitivePtr>(cnode->input(0));
if (primitive != nullptr &&
RemoveNodeInAnfExporter.count(primitive->name()) != 0) {
continue;
if (primitive != nullptr) {
if (RemoveNodeInAnfExporter.count(primitive->name()) != 0) {
continue;
}
} else {
auto primitiveT_value = GetValueNode<std::shared_ptr<PrimitiveTValue>>(cnode->input(0));
auto primT = primitiveT_value->GetPrimitiveT();
if (primT->value.type == schema::PrimitiveType_TupleGetItem ||
primT->value.type == schema::PrimitiveType_MakeTuple) {
continue;
}
}
mapRemoveGetItem_.clear();
RemoveIfMakeTuple(cnode);
RemoveIfTupleGetItem(cnode);
if (primitive != nullptr && primitive->name() == prim::kPrimReturn->name()) {
AddOutPutIfReturn(metaGraphT, cnode);
continue;
if (primitive != nullptr) {
if (primitive->name() == prim::kPrimReturn->name()) {
AddOutPutIfReturn(metaGraphT, cnode);
continue;
}
} else {
auto primitiveT_value = GetValueNode<std::shared_ptr<PrimitiveTValue>>(cnode->input(0));
auto primT = primitiveT_value->GetPrimitiveT();
if (primT->value.type == schema::PrimitiveType_Return) {
AddOutPutIfReturn(metaGraphT, cnode);
continue;
}
}
auto node = std::make_unique<schema::CNodeT>();
......@@ -134,27 +151,24 @@ schema::MetaGraphT *AnfExporter::Export(const FuncGraphPtr &funcGraph) {
primitive = GetValueNode<PrimitivePtr>(cnode->input(0));
MS_ASSERT(primitive != nullptr);
std::string opType = primitive->name();
auto nodeParser =
AnfNodePopulaterRegistry::GetInstance()->GetNodePopulater(opType);
auto nodeParser = AnfNodePopulaterRegistry::GetInstance()->GetNodePopulater(opType);
if (nodeParser == nullptr) {
MS_LOG(ERROR) << "Find op parser failed, opType: " << opType;
return nullptr;
}
std::vector<schema::TensorT *> outputs;
if (utils::isa<abstract::AbstractSequeue>(cnode->abstract())) {
auto abstract_cnode =
utils::cast<abstract::AbstractSequeuePtr>(cnode->abstract());
auto abstract_cnode = utils::cast<abstract::AbstractSequeuePtr>(cnode->abstract());
outputs.resize(abstract_cnode->size());
}
nodeParser->Parse(cnode, node.get(), &outputs);
SetOpInputNode(cnode, metaGraphT.get(), node.get());
SetOpOutputNode(outputs, metaGraphT.get(), node.get());
SetOpOutputNode(cnode, outputs, metaGraphT.get(), node.get());
metaGraphT->nodes.emplace_back(std::move(node));
continue;
}
auto primitiveT_value =
GetValueNode<std::shared_ptr<PrimitiveTValue>>(cnode->input(0));
auto primitiveT_value = GetValueNode<std::shared_ptr<PrimitiveTValue>>(cnode->input(0));
if (primitiveT_value == nullptr) {
MS_LOG(ERROR) << "PrimitiveT_value is nullptr";
return nullptr;
......@@ -166,11 +180,10 @@ schema::MetaGraphT *AnfExporter::Export(const FuncGraphPtr &funcGraph) {
return nullptr;
}
node->primitive =
std::unique_ptr<schema::PrimitiveT>(primitiveT_value->GetPrimitiveT());
node->primitive = std::unique_ptr<schema::PrimitiveT>(primitiveT_value->GetPrimitiveT());
std::vector<schema::TensorT *> outputs;
SetOpInputNode(cnode, metaGraphT.get(), node.get());
SetOpOutputNode(outputs, metaGraphT.get(), node.get());
SetOpOutputNode(cnode, outputs, metaGraphT.get(), node.get());
// add quant param
node->quantType = primitiveT_value->GetQuantType();
......@@ -244,9 +257,7 @@ schema::MetaGraphT *AnfExporter::Export(const FuncGraphPtr &funcGraph) {
return metaGraphT.release();
}
void AnfExporter::SetOpInputNode(const CNodePtr &cnode,
schema::MetaGraphT *meta_graph,
schema::CNodeT *fbNode) {
void AnfExporter::SetOpInputNode(const CNodePtr &cnode, schema::MetaGraphT *meta_graph, schema::CNodeT *fbNode) {
MS_ASSERT(nullptr != meta_graph);
MS_ASSERT(nullptr != fbNode);
if (cnode->inputs().size() <= 1) {
......@@ -281,38 +292,30 @@ void AnfExporter::SetOpInputNode(const CNodePtr &cnode,
auto paramTensor = std::make_unique<schema::TensorT>();
auto abstractBase = paramNode->abstract();
if (abstractBase == nullptr) {
MS_LOG(ERROR) << "Abstract of parameter is nullptr, "
<< paramNode->name();
MS_LOG(ERROR) << "Abstract of parameter is nullptr, " << paramNode->name();
MS_ASSERT(false);
return;
}
if (!utils::isa<abstract::AbstractTensorPtr>(abstractBase)) {
MS_LOG(ERROR) << "Abstract of parameter should be anstract tensor, "
<< paramNode->name();
MS_LOG(ERROR) << "Abstract of parameter should be anstract tensor, " << paramNode->name();
MS_ASSERT(false);
return;
}
auto abstractTensor =
utils::cast<abstract::AbstractTensorPtr>(abstractBase);
auto abstractTensor = utils::cast<abstract::AbstractTensorPtr>(abstractBase);
auto typePtr = abstractTensor->element()->GetTypeTrack();
MS_ASSERT(typePtr != nullptr);
paramTensor->dataType = typePtr->type_id();
if (!utils::isa<abstract::ShapePtr>(abstractTensor->BuildShape())) {
MS_LOG(ERROR) << "Shape of Abstract of parameter should be ShapePtr, "
<< paramNode->name();
MS_LOG(ERROR) << "Shape of Abstract of parameter should be ShapePtr, " << paramNode->name();
MS_ASSERT(false);
return;
}
paramTensor->dims =
utils::cast<abstract::ShapePtr>(abstractTensor->BuildShape())
->shape();
auto paramValue =
std::dynamic_pointer_cast<ParamValueLite>(paramNode->default_param());
paramTensor->dims = utils::cast<abstract::ShapePtr>(abstractTensor->BuildShape())->shape();
auto paramValue = std::dynamic_pointer_cast<ParamValueLite>(paramNode->default_param());
if (paramValue != nullptr) {
paramTensor->nodeType = schema::NodeType_ValueNode;
paramTensor->data.resize(paramValue->tensor_size());
memcpy(paramTensor->data.data(), paramValue->tensor_addr(),
paramValue->tensor_size());
memcpy(paramTensor->data.data(), paramValue->tensor_addr(), paramValue->tensor_size());
for (auto &ite : paramValue->quant_param()) {
auto quantPar = std::make_unique<schema::QuantParamT>();
quantPar->scale = ite->scale;
......@@ -326,8 +329,7 @@ void AnfExporter::SetOpInputNode(const CNodePtr &cnode,
paramTensor->dataType = paramValue->tensor_type();
}
}
nodeIdMap[paramNode->fullname_with_scope()] =
meta_graph->allTensors.size();
nodeIdMap[paramNode->fullname_with_scope()] = meta_graph->allTensors.size();
fbNode->inputIndex.emplace_back(meta_graph->allTensors.size());
meta_graph->allTensors.emplace_back(std::move(paramTensor));
} else if (inputNode->isa<ValueNode>()) {
......@@ -336,19 +338,15 @@ void AnfExporter::SetOpInputNode(const CNodePtr &cnode,
auto value = valueNode->value();
if (value->isa<lite::tensor::Tensor>()) {
auto valueAbstract = valueNode->abstract();
auto abstractTensor =
utils::cast<abstract::AbstractTensorPtr>(valueAbstract);
auto abstractTensor = utils::cast<abstract::AbstractTensorPtr>(valueAbstract);
auto typePtr = abstractTensor->element()->GetTypeTrack();
paramTensor->dataType = typePtr->type_id();
paramTensor->dims =
utils::cast<abstract::ShapePtr>(abstractTensor->BuildShape())
->shape();
paramTensor->dims = utils::cast<abstract::ShapePtr>(abstractTensor->BuildShape())->shape();
paramTensor->nodeType = schema::NodeType_ValueNode;
auto data = value->cast<lite::tensor::TensorPtr>();
paramTensor->data.resize(data->Size());
memcpy(paramTensor->data.data(), data->Data(), data->Size());
nodeIdMap[valueNode->fullname_with_scope()] =
meta_graph->allTensors.size();
nodeIdMap[valueNode->fullname_with_scope()] = meta_graph->allTensors.size();
fbNode->inputIndex.emplace_back(meta_graph->allTensors.size());
meta_graph->allTensors.emplace_back(std::move(paramTensor));
} else if (value->isa<mindspore::Int32Imm>()) {
......@@ -376,30 +374,44 @@ void AnfExporter::SetOpInputNode(const CNodePtr &cnode,
}
}
void AnfExporter::SetOpOutputNode(
const std::vector<schema::TensorT *> &outputTensors,
schema::MetaGraphT *graph, schema::CNodeT *cnode) {
void AnfExporter::SetOpOutputNode(const CNodePtr &cnode, const std::vector<schema::TensorT *> &outputTensors,
schema::MetaGraphT *graph, schema::CNodeT *fbnode) {
MS_ASSERT(nullptr != graph);
MS_ASSERT(nullptr != cnode);
std::string cnodeName = cnode->name;
MS_ASSERT(nullptr != fbnode);
std::string cnodeName = fbnode->name;
if (!outputTensors.empty()) {
int i = 0;
for (auto outputTensor : outputTensors) {
std::string name = cnodeName + "_o:" + std::to_string(i);
auto msTensor = new schema::TensorT();
msTensor->nodeType = schema::NodeType_Parameter;
nodeIdMap[name] = graph->allTensors.size();
cnode->outputIndex.emplace_back(graph->allTensors.size());
graph->allTensors.emplace_back(msTensor);
fbnode->outputIndex.emplace_back(graph->allTensors.size());
graph->allTensors.emplace_back(outputTensor);
i++;
}
return;
}
auto msTensor = new schema::TensorT();
msTensor->nodeType = schema::NodeType_Parameter;
cnode->outputIndex.emplace_back(graph->allTensors.size());
nodeIdMap[cnodeName] = graph->allTensors.size();
graph->allTensors.emplace_back(msTensor);
if (utils::isa<abstract::AbstractTuple>(cnode->abstract())) {
auto tuple = std::reinterpret_pointer_cast<abstract::AbstractTuple>(cnode->abstract());
for (int i = 0; i < tuple->size(); i++) {
auto msTensor = new schema::TensorT();
msTensor->nodeType = schema::NodeType_Parameter;
fbnode->outputIndex.emplace_back(graph->allTensors.size());
if (tuple->size() == 1) {
nodeIdMap[cnodeName] = graph->allTensors.size();
} else {
std::string name = cnodeName + "_o:" + std::to_string(i);
nodeIdMap[name] = graph->allTensors.size();
}
graph->allTensors.emplace_back(msTensor);
}
} else {
auto msTensor = new schema::TensorT();
msTensor->nodeType = schema::NodeType_Parameter;
fbnode->outputIndex.emplace_back(graph->allTensors.size());
nodeIdMap[cnodeName] = graph->allTensors.size();
graph->allTensors.emplace_back(msTensor);
}
}
schema::MetaGraphT *Export(const FuncGraphPtr &funcGraph) {
......
......@@ -32,8 +32,8 @@ class AnfExporter {
AnfExporter() = default;
virtual ~AnfExporter() = default;
schema::MetaGraphT *Export(const FuncGraphPtr &funcGraph);
void SetOpOutputNode(const std::vector<schema::TensorT *> &outputTensors, schema::MetaGraphT *graph,
schema::CNodeT *cnode);
void SetOpOutputNode(const CNodePtr &cnode, const std::vector<schema::TensorT *> &outputTensors,
schema::MetaGraphT *graph, schema::CNodeT *fbnode);
void SetOpInputNode(const CNodePtr &cnode, schema::MetaGraphT *meta_graph, schema::CNodeT *fbNode);
void RemoveIfMakeTuple(const CNodePtr &cnode);
bool RemoveIfTupleGetItem(const CNodePtr &cnode);
......@@ -47,4 +47,3 @@ class AnfExporter {
schema::MetaGraphT *Export(const FuncGraphPtr &funcGraph);
} // namespace mindspore::lite
#endif // MINDSPORE_LITE_SRC_ANF_EXPORTER_ANF_EXPORTER_H_
......@@ -71,11 +71,11 @@ int AnfImporterFromMetaGraphT::ConverterCNode() {
for (size_t i = 0; i < meta_graph_->nodes.size(); i++) {
auto &cNode = meta_graph_->nodes.at(i);
MS_EXCEPTION_IF_NULL(cNode);
auto tensor_id = cNode->outputIndex.front();
if (nullptr != GetNode(tensor_id)) {
continue;
}
bool flag = false;
if (cNode->outputIndex.size() > 1) {
flag = true;
}
auto primTValue = std::make_shared<PrimitiveTValue>(cNode->primitive.release());
cNode->primitive = nullptr;
auto value_node = NewValueNode(primTValue);
......@@ -90,9 +90,39 @@ int AnfImporterFromMetaGraphT::ConverterCNode() {
// todo: CheckInputNodeType, the first node should be op;
op_inputs.push_back(node);
}
auto cnode = func_graph_->NewCNode(op_inputs);
cnode->set_fullname_with_scope(cNode->name);
AddNode(tensor_id, cnode);
auto new_cnode = func_graph_->NewCNode(op_inputs);
new_cnode->set_fullname_with_scope(cNode->name);
std::vector<uint32_t> out_tensor_ids = cNode->outputIndex;
AbstractBasePtrList ptr_list;
int total = 0;
for (auto out_tensor_id : out_tensor_ids) {
if (nullptr != GetNode(out_tensor_id)) {
ptr_list.push_back(GetNode(out_tensor_id)->abstract());
continue;
}
std::vector<int> shape;
auto &tensor = meta_graph_->allTensors.at(out_tensor_id);
for (int &dim : tensor->dims) {
shape.push_back(dim);
}
auto type_id = static_cast<TypeId>(tensor->dataType);
auto type_ptr = TypeIdToType(type_id);
auto abstract_tensor = std::make_shared<abstract::AbstractTensor>(type_ptr, shape);
auto getItemPrim = NewValueNode(prim::kPrimTupleGetItem);
if (flag) {
auto getItemIndex = NewValueNode(MakeValue<int>(total++));
std::vector<AnfNodePtr> inputs{getItemPrim, new_cnode, getItemIndex};
CNodePtr new_item_cnode = func_graph_->NewCNode(inputs);
AddNode(out_tensor_id, new_item_cnode);
} else {
AddNode(out_tensor_id, new_cnode);
}
ptr_list.push_back(std::move(abstract_tensor));
}
new_cnode->set_abstract(std::make_shared<abstract::AbstractTuple>(ptr_list));
}
return RET_OK;
}
......@@ -120,4 +150,3 @@ void AnfImporterFromMetaGraphT::AddReturnCNode() {
FuncGraphPtr AnfImporterFromMetaGraphT::GetResult() { return this->func_graph_; }
} // namespace mindspore::lite
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册