p47068915
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Codecheck clearing for r1.7 parallel module.

This commit is contained in:
liuluobin 2022-03-15 18:27:08 +08:00
parent da342d8973
commit 3a85c2e002
42 changed files with 222 additions and 188 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
mindspore/ccsrc/frontend/parallel/allreduce_fusion/allreduce_graph.cc
View File

@ -167,7 +167,7 @@ void AllreduceGraph::PrintArnodeSet() const {
void AllreduceGraph::SortArnode() {
arnode_vec_.clear();
for (auto &node : arnode_set_) {
arnode_vec_.emplace_back(*node);
(void)arnode_vec_.emplace_back(*node);
}
std::sort(arnode_vec_.begin(), arnode_vec_.end(), std::greater<>());
}

4
mindspore/ccsrc/frontend/parallel/allreduce_fusion/allreduce_node.cc
View File

@ -26,7 +26,7 @@ Status AllreduceNode::AddNext(const AllreduceNodePtr &next_node) {
MS_LOG(ERROR) << "next_node is nullptr!";
return FAILED;
}
next_.emplace_back(next_node);
(void)next_.emplace_back(next_node);
return SUCCESS;
}
@ -39,7 +39,7 @@ Status AllreduceNode::AddPrev(const AllreduceNodePtr &prev_node, double dist, do
MS_LOG(ERROR) << "dist must be positive! dist: " << dist;
return FAILED;
}
prev_.emplace_back(prev_node);
(void)prev_.emplace_back(prev_node);
double add_dist = prev_node->depend_feat_size() + dist;
depend_feat_size_ += add_dist;
if (depend_feat_size_ > *max) {

4
mindspore/ccsrc/frontend/parallel/auto_parallel/costmodel.cc
View File

@ -50,7 +50,7 @@ void SimplifyForDecreasingCommunicationForward(CostPtrList *clist_ptrs) {
for (size_t i = 0; i < clist_ptrs->size(); ++i) {
if ((ret.size() == size_t(0)) ||
(clist_ptrs->at(id[i])->communication_forward_ < ret.back()->communication_forward_)) {
ret.emplace_back(std::move(clist_ptrs->at(id[i])));
(void)ret.emplace_back(std::move(clist_ptrs->at(id[i])));
}
}
*clist_ptrs = std::move(ret);
@ -73,7 +73,7 @@ void SimplifyForDecreasingCommunicationWithPartialPara(CostPtrList *clist_ptrs)
for (size_t i = 0; i < clist_ptrs->size(); ++i) {
if ((ret.size() == size_t(0)) ||
(clist_ptrs->at(id[i])->communication_with_partial_para_ < ret.back()->communication_with_partial_para_)) {
ret.emplace_back(std::move(clist_ptrs->at(id[i])));
(void)ret.emplace_back(std::move(clist_ptrs->at(id[i])));
}
}
*clist_ptrs = std::move(ret);

10
mindspore/ccsrc/frontend/parallel/auto_parallel/costmodel.h
View File

@ -125,6 +125,7 @@ struct OpEliminationDecision : public Decision {
right_cost_(std::move(r_cost)) {
type_ = DecisionType::OP_ELIMINATION;
}
~OpEliminationDecision() override = default;
StrategyPtr op_strategy_;
CostPtr left_cost_;
@ -144,6 +145,7 @@ struct EdgeEliminationDecision : public Decision {
explicit EdgeEliminationDecision(CostPtrList cost_list) : edges_cost_list_(std::move(cost_list)) {
type_ = DecisionType::EDGE_ELIMINATION;
}
~EdgeEliminationDecision() override = default;
CostPtrList edges_cost_list_;
MS_DECLARE_PARENT(EdgeEliminationDecision, Decision);
@ -167,6 +169,7 @@ struct MergeEliminationDecision : public Decision {
target_op_cost_(std::move(target_op_c)) {
type_ = DecisionType::MERGE_ELIMINATION;
}
~MergeEliminationDecision() override = default;
StrategyPtr merged_op_strategy_;
CostPtr merged_op_cost_;
@ -194,6 +197,7 @@ struct ContractEliminationDecision : public Decision {
target_cost_(std::move(tar_cost)) {
type_ = DecisionType::CONTRACT_ELIMINATION;
}
~ContractEliminationDecision() override = default;
StrategyPtr contracted_op_strategy_;
CostPtr contracted_op_cost_;
@ -228,6 +232,8 @@ struct SourceEliminationDecision : public Decision {
op2_cost_(std::move(op2_c)) {
type_ = DecisionType::SOURCE_ELIMINATION;
}
~SourceEliminationDecision() override = default;
StrategyPtr op1_strategy_;
CostPtr op1_cost_;
StrategyPtr op2_strategy_;
@ -258,6 +264,7 @@ struct TriangleEliminationDecision : public Decision {
right_node_cost_(std::move(r_node_cost)) {
type_ = DecisionType::TRIANGLE_ELIMINATION;
}
~TriangleEliminationDecision() override = default;
StrategyPtr eliminated_op_strategy_;
CostPtr eliminated_op_cost_;
@ -288,6 +295,7 @@ struct StarEliminationDecision : public Decision {
succ_ops_cost_list_(std::move(succ_ops_clist)) {
type_ = DecisionType::STAR_ELIMINATION;
}
~StarEliminationDecision() override = default;
StrategyPtr eliminated_op_strategy_;
CostPtr eliminated_op_cost_;
@ -308,6 +316,7 @@ struct FinalDecision : public Decision {
right_cost_(std::move(r_cost)) {
type_ = DecisionType::FINAL_TYPE;
}
~FinalDecision() override = default;
StrategyPtr u_strategy_;
StrategyPtr v_strategy_;
@ -323,6 +332,7 @@ struct FinalSingleDecision : public Decision {
FinalSingleDecision(StrategyPtr u_stra, CostPtr u_cost) : u_strategy_(std::move(u_stra)), u_cost_(std::move(u_cost)) {
type_ = DecisionType::FINAL_SINGLE;
}
~FinalSingleDecision() override = default;
StrategyPtr u_strategy_;
CostPtr u_cost_;

9
mindspore/ccsrc/frontend/parallel/auto_parallel/dp_algo_costmodel.h
View File

@ -44,6 +44,7 @@ namespace parallel {
struct Elimination : public Base {
enum EliminationType { OPERA, EDGE, MERGE, CONTRACT, SOURCE, TRIANGLE, STAR };
Elimination(EdgePtr n_edge, EliminationType ty) : new_edge_(std::move(n_edge)), type_(ty) {}
~Elimination() override = default;
EdgePtr new_edge_;
EliminationType type_;
@ -56,6 +57,7 @@ struct OpElimination : public Elimination {
left_edge_(std::move(l_edge)),
op_(std::move(op_info)),
right_edge_(std::move(r_edge)) {}
~OpElimination() override = default;
EdgePtr left_edge_;
OperatorInfoPtr op_;
@ -67,6 +69,7 @@ struct OpElimination : public Elimination {
struct EdgeElimination : public Elimination {
EdgeElimination(const EdgePtr &n_edge, std::vector<EdgePtr> eds)
: Elimination(n_edge, Elimination::EliminationType::EDGE), edges_(std::move(eds)) {}
~EdgeElimination() override = default;
std::vector<EdgePtr> edges_;
MS_DECLARE_PARENT(EdgeElimination, Elimination);
@ -79,6 +82,7 @@ struct MergeElimination : public Elimination {
merged_node_(std::move(u_info)),
dir_edge_(std::move(merged_target_edge)),
target_node_(std::move(v_info)) {}
~MergeElimination() override = default;
OperatorInfoPtr merged_node_;
EdgePtr dir_edge_;
@ -93,6 +97,7 @@ struct ContractElimination : public Elimination {
contracted_node_(std::move(con_info)),
dir_edge_(std::move(tar_con_edge)),
target_node_(std::move(tar_info)) {}
~ContractElimination() override = default;
OperatorInfoPtr contracted_node_;
EdgePtr dir_edge_;
@ -111,6 +116,8 @@ struct SourceElimination : public Elimination {
secondary_source_(std::move(s_source)),
secondary_succ_edges_(std::move(s_succ_edges)),
secondary_new_succ_edges_(std::move(s_new_succ_edges)) {}
~SourceElimination() override = default;
OperatorInfoPtr primary_source_;
std::vector<EdgePtr> primary_succ_edges_;
std::vector<EdgePtr> primary_new_succ_edges_;
@ -130,6 +137,7 @@ struct TriangleElimination : public Elimination {
left_node_(std::move(l_node)),
right_edge_(std::move(r_edge)),
right_node_(std::move(r_node)) {}
~TriangleElimination() override = default;
OperatorInfoPtr eliminated_node_;
EdgePtr left_edge_;
@ -146,6 +154,7 @@ struct StarElimination : public Elimination {
eliminated_node_(std::move(elimi_node)),
succ_edges_(std::move(s_edges)),
succ_ops_(std::move(s_ops)) {}
~StarElimination() override = default;
OperatorInfoPtr eliminated_node_;
std::vector<EdgePtr> succ_edges_;

12
mindspore/ccsrc/frontend/parallel/auto_parallel/edge_costmodel.cc
View File

@ -35,11 +35,11 @@ Status Edge::InitEdgeCost() {
for (auto &swc : prev_op_->GetStrategyCost()) {
MS_EXCEPTION_IF_NULL(swc);
pre_op_output_.emplace_back(std::make_pair(swc->strategy_ptr, swc->outputs_ptr));
(void)pre_op_output_.emplace_back(std::make_pair(swc->strategy_ptr, swc->outputs_ptr));
}
for (auto &swc : next_op_->GetStrategyCost()) {
MS_EXCEPTION_IF_NULL(swc);
next_op_input_.emplace_back(std::make_pair(swc->strategy_ptr, swc->inputs_ptr));
(void)next_op_input_.emplace_back(std::make_pair(swc->strategy_ptr, swc->inputs_ptr));
}
if (is_identity_edge) {
for (auto &target_output : pre_op_output_) {
@ -256,7 +256,7 @@ void Edge::CreateOpEliminationSubCostList(StrategyPtr op_strategy, const CostPtr
communication_without_para + gamma * (communication - communication_without_para);
cost->memory_with_reuse_ = memory_cost;
cost->communication_forward_ = communication_forward;
ret_cost_list->emplace_back(std::move(cost));
(void)ret_cost_list->emplace_back(std::move(cost));
}
}
}
@ -565,15 +565,15 @@ bool Edge::CheckStrategyConsistency(StrategyPtr prev_stra, StrategyPtr next_stra
return true;
}
void Edge::SetCostMapAndInputOutput(std::map<CostPtrKey, CostPtrList> &cost_map) {
void Edge::SetCostMapAndInputOutput(const std::map<CostPtrKey, CostPtrList> &cost_map) {
cost_map_ = cost_map;
pre_op_output_.clear();
next_op_input_.clear();
for (auto &key_value : cost_map_) {
auto &key_pair = key_value.first;
pre_op_output_.emplace_back(std::pair<StrategyPtr, std::vector<TensorInfo>>(key_pair.first, {}));
next_op_input_.emplace_back(std::pair<StrategyPtr, std::vector<TensorInfo>>(key_pair.second, {}));
(void)pre_op_output_.emplace_back(std::pair<StrategyPtr, std::vector<TensorInfo>>(key_pair.first, {}));
(void)next_op_input_.emplace_back(std::pair<StrategyPtr, std::vector<TensorInfo>>(key_pair.second, {}));
}
}

4
mindspore/ccsrc/frontend/parallel/auto_parallel/edge_costmodel.h
View File

@ -95,7 +95,7 @@ class Edge {
StrategyPtr GetNextOpStrategyByReshapeSWCIndex(int64_t swc_index);
bool CheckStrategyConsistency(StrategyPtr, StrategyPtr);
void SetCostMapAndInputOutput(std::map<CostPtrKey, CostPtrList> &);
void SetCostMapAndInputOutput(const std::map<CostPtrKey, CostPtrList> &);
// For two operators u--->v, given the output tensor layout of u,
// and the input tensor layout of v, return the redistribution cost,
// and the op_list to carry out the redistribution.
@ -185,7 +185,7 @@ class Edge {
int64_t is_output_critical_ = 0;
// Returns whether two double variable are equal.
bool IsDoubleEqual(double x, double y) { return std::abs(x - y) < EPS; }
bool IsDoubleEqual(double x, double y) const { return std::abs(x - y) < EPS; }
};
} // namespace parallel
} // namespace mindspore

4
mindspore/ccsrc/frontend/parallel/auto_parallel/graph_costmodel.h
View File

@ -57,8 +57,8 @@ class CostGraph {
std::map<OperatorInfoPtr, bool> *);
// the edge is in the form: u --> v
void AddEdge(OperatorInfoPtr u_node, OperatorInfoPtr v_node, const EdgePtr &edge);
std::vector<std::shared_ptr<Edge>> GetOriginalPrevEdges(OperatorInfoPtr v_node) { return in_edges_[v_node]; }
std::vector<std::shared_ptr<Edge>> GetOriginalNextEdges(OperatorInfoPtr u_node) { return out_edges_[u_node]; }
std::vector<std::shared_ptr<Edge>> GetOriginalPrevEdges(const OperatorInfoPtr &v_node) { return in_edges_[v_node]; }
std::vector<std::shared_ptr<Edge>> GetOriginalNextEdges(const OperatorInfoPtr &u_node) { return out_edges_[u_node]; }
// An edge is uniquely identified by its name, and its output index and input index.
bool IsEdgeInCostGraph(const std::string &, size_t, size_t);

22
mindspore/ccsrc/frontend/parallel/auto_parallel/operator_costmodel.cc
View File

@ -1802,7 +1802,7 @@ void MatmulDDSCost::CalculateInputsInMemory(const std::map<size_t, bool> &) {
}
double CropAndResizeCost::GetForwardCommCost(const std::vector<TensorInfo> &inputs,
const std::vector<TensorInfo> &outputs, int64_t stage_id) const {
const std::vector<TensorInfo> &outputs, int64_t) const {
double result = 0.0;
if (outputs_type_lengths_.size() != outputs.size()) {
MS_LOG(EXCEPTION) << "Invalid inputs type size " << inputs_type_lengths_.size() << " for CropAndResize cost.";
@ -1821,8 +1821,8 @@ double CropAndResizeCost::GetForwardCommCost(const std::vector<TensorInfo> &inpu
return result;
}
double CropAndResizeCost::GetBackwardCommCost(const std::vector<TensorInfo> &inputs,
const std::vector<TensorInfo> &outputs, int64_t stage_id) const {
double CropAndResizeCost::GetBackwardCommCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &,
int64_t stage_id) const {
double result = 0.0;
CheckGlobalDeviceManager();
MS_EXCEPTION_IF_NULL(g_device_manager);
@ -1847,7 +1847,7 @@ double CropAndResizeCost::GetBackwardCommCost(const std::vector<TensorInfo> &inp
}
double CropAndResizeCost::GetForwardComputationCost(const std::vector<TensorInfo> &inputs,
const std::vector<TensorInfo> &outputs, int64_t stage_id) const {
const std::vector<TensorInfo> &, int64_t) const {
if (inputs_type_lengths_.size() != inputs.size()) {
MS_LOG(EXCEPTION) << "Invalid inputs type size " << inputs_type_lengths_.size() << " for CropAndResize cost.";
}
@ -1872,8 +1872,8 @@ double CropAndResizeCost::GetForwardComputationCost(const std::vector<TensorInfo
return result;
}
double CropAndResizeCost::GetBackwardComputationCost(const std::vector<TensorInfo> &inputs,
const std::vector<TensorInfo> &outputs, int64_t stage_id) const {
double CropAndResizeCost::GetBackwardComputationCost(const std::vector<TensorInfo> &,
const std::vector<TensorInfo> &outputs, int64_t) const {
Shape output0_slice_shape = outputs[0].slice_shape();
double result = 0.0;
// don't split batch
@ -1910,7 +1910,7 @@ void CropAndResizeCost::CalculateInputsInMemory(const std::map<size_t, bool> &pr
}
double ROIAlignCost::GetForwardCommCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
int64_t stage_id) const {
int64_t) const {
double result = 0.0;
if (outputs_type_lengths_.size() != outputs.size()) {
MS_LOG(EXCEPTION) << "Invalid inputs type size " << inputs_type_lengths_.size() << " for CropAndResize cost.";
@ -1929,8 +1929,8 @@ double ROIAlignCost::GetForwardCommCost(const std::vector<TensorInfo> &inputs, c
return result;
}
double ROIAlignCost::GetForwardComputationCost(const std::vector<TensorInfo> &inputs,
const std::vector<TensorInfo> &outputs, int64_t stage_id) const {
double ROIAlignCost::GetForwardComputationCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &,
int64_t) const {
if (inputs_type_lengths_.size() != inputs.size()) {
MS_LOG(EXCEPTION) << "Invalid inputs type size " << inputs_type_lengths_.size() << " for CropAndResize cost.";
}
@ -1951,8 +1951,8 @@ double ROIAlignCost::GetForwardComputationCost(const std::vector<TensorInfo> &in
return result;
}
double ROIAlignCost::GetBackwardComputationCost(const std::vector<TensorInfo> &inputs,
const std::vector<TensorInfo> &outputs, int64_t stage_id) const {
double ROIAlignCost::GetBackwardComputationCost(const std::vector<TensorInfo> &, const std::vector<TensorInfo> &outputs,
int64_t) const {
Shape output0_slice_shape = outputs[0].slice_shape();
double result = 0.0;
// don't split batch

45
mindspore/ccsrc/frontend/parallel/auto_parallel/operator_costmodel.h
View File

@ -114,7 +114,6 @@ class OperatorCost {
// in the inference phase.
int64_t is_outputs_critical_ = -1;
};
using OperatorCostPtr = std::shared_ptr<OperatorCost>;
class MatMulCost : public OperatorCost {
public:
@ -1016,13 +1015,11 @@ class UniformCandidateSamplerCost : public OperatorCost {
int64_t stage_id) const override {
return GetForwardCommCost(inputs, outputs, stage_id) + GetBackwardCommCost(inputs, outputs, stage_id);
}
double GetForwardCommCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
int64_t stage_id) const override {
return 0;
double GetForwardCommCost(const std::vector<TensorInfo> &, const std::vector<TensorInfo> &, int64_t) const override {
return 0.0;
}
double GetBackwardCommCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
int64_t stage_id) const override {
return 0;
double GetBackwardCommCost(const std::vector<TensorInfo> &, const std::vector<TensorInfo> &, int64_t) const override {
return 0.0;
}
double GetComputationCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
int64_t stage_id) const override {
@ -1030,7 +1027,7 @@ class UniformCandidateSamplerCost : public OperatorCost {
}
double GetForwardComputationCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
int64_t stage_id) const override;
double GetBackwardComputationCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
double GetBackwardComputationCost(const std::vector<TensorInfo> &, const std::vector<TensorInfo> &,
int64_t) const override {
return 0.0;
}
@ -1106,12 +1103,10 @@ class MatmulDDSCost : public OperatorCost {
int64_t stage_id) const override {
return GetForwardCommCost(inputs, outputs, stage_id) + GetBackwardCommCost(inputs, outputs, stage_id);
}
double GetForwardCommCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
int64_t stage_id) const override {
double GetForwardCommCost(const std::vector<TensorInfo> &, const std::vector<TensorInfo> &, int64_t) const override {
return 0.0;
};
double GetBackwardCommCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
int64_t stage_id) const override {
double GetBackwardCommCost(const std::vector<TensorInfo> &, const std::vector<TensorInfo> &, int64_t) const override {
return 0.0;
};
@ -1122,8 +1117,8 @@ class MatmulDDSCost : public OperatorCost {
}
double GetForwardComputationCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
int64_t stage_id) const override;
double GetBackwardComputationCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
int64_t stage_id) const override {
double GetBackwardComputationCost(const std::vector<TensorInfo> &, const std::vector<TensorInfo> &,
int64_t) const override {
return 0.0;
};
// Not taking account of output
@ -1143,17 +1138,17 @@ class CropAndResizeCost : public OperatorCost {
return GetForwardCommCost(inputs, outputs, stage_id) + GetBackwardCommCost(inputs, outputs, stage_id);
}
double GetForwardCommCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
int64_t stage_id) const override;
int64_t) const override;
double GetBackwardCommCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
int64_t stage_id) const override;
double GetComputationCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
int64_t stage_id) const override {
return GetForwardComputationCost(inputs, outputs, stage_id) + GetBackwardComputationCost(inputs, outputs, stage_id);
}
double GetForwardComputationCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
int64_t stage_id) const override;
double GetBackwardComputationCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
int64_t stage_id) const override;
double GetForwardComputationCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &,
int64_t) const override;
double GetBackwardComputationCost(const std::vector<TensorInfo> &, const std::vector<TensorInfo> &,
int64_t) const override;
// Taking account for input
void CalculateInputsInMemory(const std::map<size_t, bool> &prev_output_in_mem) override;
// Not taking account of output
@ -1179,17 +1174,17 @@ class ROIAlignCost : public CropAndResizeCost {
~ROIAlignCost() override = default;
double GetForwardCommCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
int64_t stage_id) const override;
double GetForwardComputationCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
int64_t stage_id) const override;
double GetBackwardComputationCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &outputs,
int64_t stage_id) const override;
int64_t) const override;
double GetForwardComputationCost(const std::vector<TensorInfo> &inputs, const std::vector<TensorInfo> &,
int64_t) const override;
double GetBackwardComputationCost(const std::vector<TensorInfo> &, const std::vector<TensorInfo> &outputs,
int64_t) const override;
// Taking account for input
void CalculateInputsInMemory(const std::map<size_t, bool> &prev_output_in_mem) override;
// Taking account of output
void CalculateOutputInMemory() override;
void set_pooled_shape(Shape pooled_shape) { pooled_shape_ = pooled_shape; }
void set_pooled_shape(const Shape &pooled_shape) { pooled_shape_ = pooled_shape; }
protected:
Shape pooled_shape_;

55
mindspore/ccsrc/frontend/parallel/auto_parallel/rec_core/rec_cost.cc
View File

@ -140,30 +140,26 @@ StrategyRec CostMatMul::GetOptimalStr(const Graph::NodeType &node,
static_cast<int64_t>(node.apply.arguments[0].tensor_shape.shape_w * node.apply.arguments[0].tensor_str.str_w);
std::vector<double> cost_op;
std::vector<std::vector<float>> mode;
if (edge_i < 2 || edge_i % 2 != 0) {
cost_op.push_back(DOUBLE_MAX);
} else {
cost_op.push_back(StrConcatDimI(edge_j, edge_k) + CostRedis(node, node_name_to_strategy,
mode = {{1, 1, 0.5, 1}, {1, 1, 1, 1}, {1, 1, 0.5, 1}},
graph));
std::vector<std::vector<float>> mode = {{1, 1, 0.5, 1}, {1, 1, 1, 1}, {1, 1, 0.5, 1}};
cost_op.push_back(StrConcatDimI(edge_j, edge_k) + CostRedis(node, node_name_to_strategy, mode, graph));
}
if (edge_j < 2 || edge_j % 2 != 0) {
cost_op.push_back(DOUBLE_MAX);
} else {
cost_op.push_back(StrConcatDimJ(edge_i, edge_k) + CostRedis(node, node_name_to_strategy,
mode = {{1, 1, 1, 1}, {1, 1, 1, 0.5}, {1, 1, 1, 0.5}},
graph));
std::vector<std::vector<float>> mode = {{1, 1, 1, 1}, {1, 1, 1, 0.5}, {1, 1, 1, 0.5}};
cost_op.push_back(StrConcatDimJ(edge_i, edge_k) + CostRedis(node, node_name_to_strategy, mode, graph));
}
if (edge_k < 2 || edge_k % 2 != 0) {
cost_op.push_back(DOUBLE_MAX);
} else {
cost_op.push_back(StrReduceDimK(edge_i, edge_j) + CostRedis(node, node_name_to_strategy,
mode = {{1, 1, 1, 0.5}, {1, 1, 0.5, 1}, {1, 1, 1, 1}},
graph));
std::vector<std::vector<float>> mode = {{1, 1, 1, 0.5}, {1, 1, 0.5, 1}, {1, 1, 1, 1}};
cost_op.push_back(StrReduceDimK(edge_i, edge_j) + CostRedis(node, node_name_to_strategy, mode, graph));
}
return ChoseStr(cost_op, node.apply.str);
@ -260,13 +256,12 @@ StrategyRec CostConvolution::GetOptimalStr(
std::vector<double> cost_op;
cost_op.reserve(7);
std::vector<std::vector<float>> mode;
if (input_tensor_n < 2 || input_tensor_n % 2 != 0) {
cost_op.push_back(DOUBLE_MAX);
} else {
cost_op.push_back(StrDimB(tensor_filter) + CostRedis(node, node_name_to_strategy,
mode = {{0.5, 1, 1, 1}, {1, 1, 1, 1}, {0.5, 1, 1, 1}}, graph));
std::vector<std::vector<float>> mode = {{0.5, 1, 1, 1}, {1, 1, 1, 1}, {0.5, 1, 1, 1}};
cost_op.push_back(StrDimB(tensor_filter) + CostRedis(node, node_name_to_strategy, mode, graph));
}
cost_op.push_back(DOUBLE_MAX);
@ -275,8 +270,8 @@ StrategyRec CostConvolution::GetOptimalStr(
if (channel_partition == false || tensor_filter < 2 || tensor_filter % 2 != 0) {
cost_op.push_back(DOUBLE_MAX);
} else {
cost_op.push_back(StrDimK(tensor_in) + CostRedis(node, node_name_to_strategy,
mode = {{1, 1, 1, 1}, {0.5, 1, 1, 1}, {1, 0.5, 1, 1}}, graph));
std::vector<std::vector<float>> mode = {{1, 1, 1, 1}, {0.5, 1, 1, 1}, {1, 0.5, 1, 1}};
cost_op.push_back(StrDimK(tensor_in) + CostRedis(node, node_name_to_strategy, mode, graph));
}
cost_op.push_back(DOUBLE_MAX);
@ -285,8 +280,8 @@ StrategyRec CostConvolution::GetOptimalStr(
if (channel_partition == false || tensor_filter_c < 2 || tensor_filter_c % 2 != 0) {
cost_op.push_back(DOUBLE_MAX);
} else {
cost_op.push_back(StrDimQ(tensor_out) + CostRedis(node, node_name_to_strategy,
mode = {{1, 0.5, 1, 1}, {1, 0.5, 1, 1}, {1, 1, 1, 1}}, graph));
std::vector<std::vector<float>> mode = {{1, 0.5, 1, 1}, {1, 0.5, 1, 1}, {1, 1, 1, 1}};
cost_op.push_back(StrDimQ(tensor_out) + CostRedis(node, node_name_to_strategy, mode, graph));
}
return ChoseStr(cost_op, node.apply.str);
@ -391,20 +386,19 @@ StrategyRec CostPooling::GetOptimalStr(const Graph::NodeType &node,
int64_t tensor_c = static_cast<int64_t>(node.tensor_parm.tensor_shape.shape_c * node.tensor_parm.tensor_str.str_c);
std::vector<double> cost_op;
std::vector<std::vector<float>> mode;
if (tensor_n < 2 || tensor_n % 2 != 0) {
cost_op.push_back(DOUBLE_MAX);
} else {
cost_op.push_back(cost_in_ + CostRedis(node, node_name_to_strategy,
mode = {{0.5, 1, 1, 1}, {0.5, 1, 1, 1}, {0.5, 1, 1, 1}}, graph));
std::vector<std::vector<float>> mode = {{0.5, 1, 1, 1}, {0.5, 1, 1, 1}, {0.5, 1, 1, 1}};
cost_op.push_back(cost_in_ + CostRedis(node, node_name_to_strategy, mode, graph));
}
if (tensor_c < 2 || tensor_c % 2 != 0) {
cost_op.push_back(DOUBLE_MAX);
} else {
cost_op.push_back(cost_in_ + CostRedis(node, node_name_to_strategy,
mode = {{1, 0.5, 1, 1}, {1, 0.5, 1, 1}, {1, 0.5, 1, 1}}, graph));
std::vector<std::vector<float>> mode = {{1, 0.5, 1, 1}, {1, 0.5, 1, 1}, {1, 0.5, 1, 1}};
cost_op.push_back(cost_in_ + CostRedis(node, node_name_to_strategy, mode, graph));
}
cost_op.push_back(DOUBLE_MAX);
@ -560,34 +554,33 @@ StrategyRec CostCommon::GetOptimalStr(const Graph::NodeType &node,
int64_t tensor_w = static_cast<int64_t>(op.arguments[0].tensor_shape.shape_w * op.arguments[0].tensor_str.str_w);
std::vector<double> cost_op;
std::vector<std::vector<float>> mode;
if (tensor_n < 2 || tensor_n % 2 != 0) {
cost_op.push_back(DOUBLE_MAX);
} else {
cost_op.push_back(cost_in_ + CostRedis(node, node_name_to_strategy,
mode = {{0.5, 1, 1, 1}, {0.5, 1, 1, 1}, {0.5, 1, 1, 1}}, graph));
std::vector<std::vector<float>> mode = {{0.5, 1, 1, 1}, {0.5, 1, 1, 1}, {0.5, 1, 1, 1}};
cost_op.push_back(cost_in_ + CostRedis(node, node_name_to_strategy, mode, graph));
}
if (tensor_c < 2 || tensor_c % 2 != 0) {
cost_op.push_back(DOUBLE_MAX);
} else {
cost_op.push_back(cost_in_ + CostRedis(node, node_name_to_strategy,
mode = {{1, 0.5, 1, 1}, {1, 0.5, 1, 1}, {1, 0.5, 1, 1}}, graph));
std::vector<std::vector<float>> mode = {{1, 0.5, 1, 1}, {1, 0.5, 1, 1}, {1, 0.5, 1, 1}};
cost_op.push_back(cost_in_ + CostRedis(node, node_name_to_strategy, mode, graph));
}
if (tensor_h < 2 || tensor_h % 2 != 0) {
cost_op.push_back(DOUBLE_MAX);
} else {
cost_op.push_back(cost_in_ + CostRedis(node, node_name_to_strategy,
mode = {{1, 1, 0.5, 1}, {1, 1, 0.5, 1}, {1, 1, 0.5, 1}}, graph));
std::vector<std::vector<float>> mode = {{1, 1, 0.5, 1}, {1, 1, 0.5, 1}, {1, 1, 0.5, 1}};
cost_op.push_back(cost_in_ + CostRedis(node, node_name_to_strategy, mode, graph));
}
if (tensor_w < 2 || tensor_w % 2 != 0) {
cost_op.push_back(DOUBLE_MAX);
} else {
cost_op.push_back(cost_in_ + CostRedis(node, node_name_to_strategy,
mode = {{1, 1, 1, 0.5}, {1, 1, 1, 0.5}, {1, 1, 1, 0.5}}, graph));
std::vector<std::vector<float>> mode = {{1, 1, 1, 0.5}, {1, 1, 1, 0.5}, {1, 1, 1, 0.5}};
cost_op.push_back(cost_in_ + CostRedis(node, node_name_to_strategy, mode, graph));
}
return ChoseStr(cost_op, node.apply.str);

14
mindspore/ccsrc/frontend/parallel/auto_parallel/rec_core/rec_cost.h
View File

@ -175,6 +175,8 @@ class CostReshape {
// class CostCommon is used to compute the cost of an element-wise operator
class CostCommon {
public:
virtual ~CostCommon() = default;
virtual StrategyRec GetOptimalStr(const Graph::NodeType &node,
const std::vector<std::pair<std::string, StrategyRec>> &node_name_to_strategy,
const Graph &graph);
@ -189,11 +191,14 @@ class CostCommon {
// class CostBiasAdd is used to compute the cost of the addition between a tensor and a bias
class CostBiasAdd : public CostCommon {
StrategyRec ChoseStr(const std::vector<double> &cost_op, StrategyRec str);
protected:
StrategyRec ChoseStr(const std::vector<double> &cost_op, StrategyRec str) override;
};
// class CostAdd is used to compute the cost of Add operator.
class CostTensorAdd : public CostCommon {
StrategyRec ChoseStr(const std::vector<double> &cost_op, StrategyRec str);
protected:
StrategyRec ChoseStr(const std::vector<double> &cost_op, StrategyRec str) override;
};
// all the following operation are element-wise and have the same cost
@ -210,6 +215,8 @@ class CostCast : public CostCommon {};
// class BatchParallel is used to compute the cost of BatchParallel operator.
class CostBatchParallel {
public:
virtual ~CostBatchParallel() = default;
virtual StrategyRec GetOptimalStr(const Graph::NodeType &node);
virtual double GetMaxCostIn() const { return DOUBLE_MAX; }
@ -226,7 +233,8 @@ class CostPRelu : public CostBatchParallel {};
class CostSoftmax : public CostBatchParallel {};
class CostSoftmaxCrossEntropyWithLogits : public CostBatchParallel {
StrategyRec ChoseStr(const std::vector<double> &cost_op, StrategyRec str);
protected:
StrategyRec ChoseStr(const std::vector<double> &cost_op, StrategyRec str) override;
};
} // namespace parallel
} // namespace mindspore

9
mindspore/ccsrc/frontend/parallel/auto_parallel/rec_core/rec_parse_graph.cc
View File

@ -196,7 +196,8 @@ void Eliminate_Aux(const size_t node_index, const std::shared_ptr<Graph> &graph,
auto it = find(incoming_outputs->begin(), incoming_outputs->end(), node_index);
if (it != incoming_outputs->end()) {
it = incoming_outputs->erase(it);
incoming_outputs->insert(it, graph->nodes[node_index].node_out.begin(), graph->nodes[node_index].node_out.end());
(void)incoming_outputs->insert(it, graph->nodes[node_index].node_out.begin(),
graph->nodes[node_index].node_out.end());
}
}
@ -205,8 +206,8 @@ void Eliminate_Aux(const size_t node_index, const std::shared_ptr<Graph> &graph,
auto it = find(aux_incoming_outputs->begin(), aux_incoming_outputs->end(), node_index);
if (it != aux_incoming_outputs->end()) {
it = aux_incoming_outputs->erase(it);
aux_incoming_outputs->insert(it, graph->nodes[node_index].node_out.begin(),
graph->nodes[node_index].node_out.end());
(void)aux_incoming_outputs->insert(it, graph->nodes[node_index].node_out.begin(),
graph->nodes[node_index].node_out.end());
}
}
@ -225,7 +226,7 @@ void Eliminate_Aux(const size_t node_index, const std::shared_ptr<Graph> &graph,
graph->nodes[node_index].node_in_aux[j]);
}
} else {
outgoing_inputs->erase(it);
(void)outgoing_inputs->erase(it);
}
}
}

6
mindspore/ccsrc/frontend/parallel/auto_parallel/rec_core/rec_partition.cc
View File

@ -261,9 +261,8 @@ Status PartitionForAllDevices(const size_t num_device, const double device_memor
if (DevicesMemoryControl(num_device, device_memory, graph) != SUCCESS) {
return FAILED;
} else {
return SUCCESS;
}
return SUCCESS;
}
// Apply OP Strategy to Tensor Strategy
@ -294,7 +293,7 @@ Status DevicesMemoryControl(const size_t num_device, const double device_memory,
double used_memory = 0.0;
for (uint64_t i_node = 0; i_node < iter_nodes; i_node++) {
if (graph->nodes[i_node].info == 0) {
if (graph->nodes[i_node].info == InfoType::kApplication) {
Graph::NodeType &Node = graph->nodes[i_node];
for (int64_t index = 0; index < 2; index++) {
used_memory += Node.apply.arguments[index].tensor_str.str_n * Node.apply.arguments[index].tensor_shape.shape_n *
@ -308,7 +307,6 @@ Status DevicesMemoryControl(const size_t num_device, const double device_memory,
if (device_memory < (used_memory / num_device)) {
MS_LOG(EXCEPTION) << "Failure: Out of memory!";
return FAILED;
} else {
return SUCCESS;
}

2
mindspore/ccsrc/frontend/parallel/graph_util/generate_graph.h
View File

@ -48,7 +48,7 @@ std::string HashInstanceName(const std::string &name);
class GenerateGraph {
public:
explicit GenerateGraph(mindspore::HashMap<std::string, ValuePtr> origin_attrs)
explicit GenerateGraph(const mindspore::HashMap<std::string, ValuePtr> &origin_attrs)
: name_idx_(0), origin_attrs_(origin_attrs) {}
Status Init(const CNodePtr &cnode);
~GenerateGraph() = default;

6
mindspore/ccsrc/frontend/parallel/graph_util/graph_splitter.cc
View File

@ -72,7 +72,7 @@ void GraphSplitter::DyeGraph() {
MS_EXCEPTION_IF_NULL(func_graph_);
std::vector<AnfNodePtr> all_nodes = DeepScopedGraphSearch(func_graph_->get_return());
(void)std::for_each(all_nodes.begin(), all_nodes.end(), [this](AnfNodePtr &node) {
(void)std::for_each(all_nodes.begin(), all_nodes.end(), [this](const AnfNodePtr &node) {
MS_EXCEPTION_IF_NULL(node);
// Mark all nodes with original label at the beginning.
node_labels_[node] = default_label_;
@ -320,8 +320,8 @@ CNodePtr GraphSplitter::GenerateRecvNode(const AnfNodePtr &input, const AnfNodeP
if (input->isa<CNode>() && common::AnfAlgo::HasNodeAttr(kAttrUpdateParameter, input->cast<CNodePtr>()) &&
common::AnfAlgo::HasNodeAttr(kAttrParameterInputIndex, input->cast<CNodePtr>())) {
int64_t parameter_index = common::AnfAlgo::GetNodeAttr<int64_t>(input, kAttrParameterInputIndex);
auto kernel_with_index =
common::AnfAlgo::VisitKernel(common::AnfAlgo::GetInputNode(input->cast<CNodePtr>(), parameter_index), 0);
auto kernel_with_index = common::AnfAlgo::VisitKernel(
common::AnfAlgo::GetInputNode(input->cast<CNodePtr>(), LongToSize(parameter_index)), 0);
auto param_node = kernel_with_index.first;
recv_inputs.push_back(param_node);
recv_node_abs = param_node->abstract();

4
mindspore/ccsrc/frontend/parallel/ops_info/activation_info.cc
View File

@ -575,7 +575,9 @@ Status SqueezeInfo::GetAttrs() {
MS_EXCEPTION_IF_NULL(iter->second);
auto value_tuple = iter->second->cast<ValueTuplePtr>();
MS_EXCEPTION_IF_NULL(value_tuple);
InferAxis(value_tuple);
if (InferAxis(value_tuple) != SUCCESS) {
return FAILED;
}
attrs_[AXIS] = axis_;
return SUCCESS;
}

10
mindspore/ccsrc/frontend/parallel/ops_info/activation_info.h
View File

@ -32,7 +32,7 @@ namespace parallel {
class ActivationBase : public OperatorInfo {
public:
ActivationBase(const std::string &operator_name, const Shapes &inputs_shape, const Shapes &outputs_shape,
const PrimitiveAttrs &attrs, OperatorCostPtr cost)
const PrimitiveAttrs &attrs, const OperatorCostPtr &cost)
: OperatorInfo(operator_name, inputs_shape, outputs_shape, attrs, cost) {}
~ActivationBase() override = default;
@ -46,7 +46,7 @@ class ActivationBase : public OperatorInfo {
class Activation : public ActivationBase {
public:
Activation(const std::string &name, const Shapes &inputs_shape, const Shapes &outputs_shape,
const PrimitiveAttrs &attrs, OperatorCostPtr cost)
const PrimitiveAttrs &attrs, const OperatorCostPtr &cost)
: ActivationBase(name, inputs_shape, outputs_shape, attrs, cost) {}
~Activation() override = default;
std::vector<StrategyPtr> GenerateOpStrategies(int64_t stage_id) override;
@ -70,7 +70,7 @@ class ActivationInfo : public Activation {
class ActivationOther : public Activation {
public:
ActivationOther(const std::string &name, const Shapes &inputs_shape, const Shapes &outputs_shape,
const PrimitiveAttrs &attrs, OperatorCostPtr cost)
const PrimitiveAttrs &attrs, const OperatorCostPtr &cost)
: Activation(name, inputs_shape, outputs_shape, attrs, cost) {}
~ActivationOther() override = default;
@ -294,8 +294,8 @@ class DropoutInfo : public ActivationOther {
private:
int64_t seed0_ = 0;
int64_t seed1_ = 0;
int64_t get_seed() {
static int64_t SEED_NUM;
int64_t get_seed() const {
static int64_t SEED_NUM = 0;
return ++SEED_NUM;
}
};

2
mindspore/ccsrc/frontend/parallel/ops_info/arithmetic_info.h
View File

@ -32,7 +32,7 @@ namespace parallel {
class ArithmeticBase : public OperatorInfo {
public:
ArithmeticBase(const std::string &operator_name, const Shapes &inputs_shape, const Shapes &outputs_shape,
const PrimitiveAttrs &attrs, OperatorCostPtr cost)
const PrimitiveAttrs &attrs, const OperatorCostPtr &cost)
: OperatorInfo(operator_name, inputs_shape, outputs_shape, attrs, cost) {}
~ArithmeticBase() override = default;
std::vector<StrategyPtr> GenerateOpStrategies(int64_t) override;

6
mindspore/ccsrc/frontend/parallel/ops_info/bounding_box_encode_info.cc
View File

@ -59,9 +59,9 @@ Status BoundingBoxEncodeInfo::InferTensorMap() {
inputs_tensor_map_.clear();
outputs_tensor_map_.clear();
inputs_tensor_map_.emplace_back(TensorMap({0, -1}));
inputs_tensor_map_.emplace_back(TensorMap({0, -1}));
outputs_tensor_map_.emplace_back(TensorMap({0, -1}));
(void)inputs_tensor_map_.emplace_back(TensorMap({0, -1}));
(void)inputs_tensor_map_.emplace_back(TensorMap({0, -1}));
(void)outputs_tensor_map_.emplace_back(TensorMap({0, -1}));
return SUCCESS;
}

14
mindspore/ccsrc/frontend/parallel/ops_info/crop_and_resize_info.cc
View File

@ -57,8 +57,9 @@ Status CropAndResizeInfo::CheckStrategy(const StrategyPtr &strategy) {
Status CropAndResizeInfo::InferDevMatrixShape() {
dev_matrix_shape_.clear();
auto x_strategy = strategy_->GetInputDim().at(0);
auto boxes_strategy = strategy_->GetInputDim().at(1);
auto strategies = strategy_->GetInputDim();
auto x_strategy = strategies.at(0);
auto boxes_strategy = strategies.at(1);
dev_matrix_shape_ = {x_strategy[0], x_strategy[3], boxes_strategy[0]};
return SUCCESS;
}
@ -83,7 +84,8 @@ Status CropAndResizeInfo::InferBias() {
CheckGlobalDeviceManager();
int64_t rank = g_device_manager->rank_index_in_stage();
auto x_strategy = strategy_->GetInputDim().at(0);
auto strategies = strategy_->GetInputDim();
auto x_strategy = strategies.at(0);
Shape x_shape = inputs_shape_.at(0);
if (x_shape[0] % x_strategy[0] != 0) {
return FAILED;
@ -181,7 +183,8 @@ Status CropAndResizeInfo::InitForCostModel(const StrategyPtr &strategy, const St
constexpr size_t CROP_SIZE_INDEX = 3;
auto crop_size = GetValue<std::vector<int64_t>>(input_value_[CROP_SIZE_INDEX]);
auto x_strategy = strategy_->GetInputDim().at(0);
auto strategies = strategy_->GetInputDim();
auto x_strategy = strategies.at(0);
auto crop_and_resize_cost = std::dynamic_pointer_cast<CropAndResizeCost>(operator_cost());
crop_and_resize_cost->set_crop_size(crop_size);
crop_and_resize_cost->set_strategy(x_strategy);
@ -190,7 +193,8 @@ Status CropAndResizeInfo::InitForCostModel(const StrategyPtr &strategy, const St
}
ReplaceGraphPtr CropAndResizeInfo::replace_graph(const CNodePtr &cnode) {
auto x_strategy = strategy_->GetInputDim().at(0);
auto strategies = strategy_->GetInputDim();
auto x_strategy = strategies.at(0);
if (x_strategy[0] != 1 && ComputeReplaceGraph(cnode) != SUCCESS) {
MS_LOG(EXCEPTION) << name_ << ": ComputeReplaceGraph failed.";
}

4
mindspore/ccsrc/frontend/parallel/ops_info/crop_and_resize_info.h
View File

@ -52,8 +52,8 @@ class CropAndResizeInfo : public OperatorInfo {
Status InferGroup();
Status ComputeReplaceGraph(const CNodePtr &cnode);
int64_t slice_size_;
int64_t bias_;
int64_t slice_size_ = 0;
int64_t bias_ = 0;
Group group_;
};
} // namespace parallel

9
mindspore/ccsrc/frontend/parallel/ops_info/iou_info.cc
View File

@ -40,8 +40,7 @@ Status IOUInfo::InferDevMatrixShape() {
int64_t dev0 = strategise[1][0];
dev_matrix_shape_.clear();
dev_matrix_shape_.push_back(dev1);
dev_matrix_shape_.push_back(dev0);
dev_matrix_shape_ = {dev1, dev0};
MS_LOG(INFO) << name_ << ": The dev matrix is " << ShapeToString(dev_matrix_shape_);
return SUCCESS;
}
@ -50,9 +49,9 @@ Status IOUInfo::InferTensorMap() {
inputs_tensor_map_.clear();
outputs_tensor_map_.clear();
inputs_tensor_map_.emplace_back(TensorMap({1, -1}));
inputs_tensor_map_.emplace_back(TensorMap({0, -1}));
outputs_tensor_map_.emplace_back(TensorMap({0, 1}));
(void)inputs_tensor_map_.emplace_back(TensorMap({1, -1}));
(void)inputs_tensor_map_.emplace_back(TensorMap({0, -1}));
(void)outputs_tensor_map_.emplace_back(TensorMap({0, 1}));
return SUCCESS;
}

13
mindspore/ccsrc/frontend/parallel/ops_info/matmul_info.cc
View File

@ -449,7 +449,8 @@ Status MatMulBase::GenerateStrategiesBase(int64_t stage_id, size_t dev_num, cons
};
recursive(0, dev_num);
if (sp_vector->empty()) {
MS_LOG(EXCEPTION) << name_ << " : No available strategy.";
MS_LOG(ERROR) << name_ << " : No available strategy.";
return FAILED;
}
return Status::SUCCESS;
}
@ -472,9 +473,10 @@ Status MatMulBase::GenerateStrategiesNotPower2(int64_t stage_id, size_t dev_num_
continue;
}
auto new_stra_arrays{stra_arrays};
new_stra_arrays[i][j] = new_stra_arrays[i][j] * dev_num_not_2_power;
new_stra_arrays[i][j] = new_stra_arrays[i][j] * SizeToLong(dev_num_not_2_power);
if (i == 0 && j == related_dim_left) {
new_stra_arrays[1][related_dim_right] = new_stra_arrays[1][related_dim_right] * dev_num_not_2_power;
new_stra_arrays[1][related_dim_right] =
new_stra_arrays[1][related_dim_right] * SizeToLong(dev_num_not_2_power);
}
StrategyPtr new_stra = std::make_shared<Strategy>(stage_id, new_stra_arrays);
sp_vector.push_back(new_stra);
@ -517,6 +519,7 @@ Status MatMulBase::GenerateStrategies(int64_t stage_id) {
std::vector<StrategyPtr> sp_vector_2_power_part;
if (dev_num_2_power == 0) {
if (GenerateStrategiesBase(stage_id, dev_num, input0_shape, input1_shape, &sp_vector_2_power_part) != SUCCESS) {
MS_LOG(ERROR) << "No available strategy.";
return FAILED;
}
strategy_cost_.clear();
@ -660,7 +663,7 @@ Status MatMulBase::CheckForTensorSliceValid() const {
std::shared_ptr<Strategys> BatchMatMulInfo::GenerateBatchStrategies() {
Dimensions batch_strategy(inputs_shape_[1].size() - 1, 1);
batch_strategy.insert(batch_strategy.begin(), stage_device_size_);
(void)batch_strategy.insert(batch_strategy.begin(), stage_device_size_);
Strategys strategy_v = {batch_strategy, batch_strategy};
return std::make_shared<Strategys>(strategy_v);
}
@ -706,7 +709,7 @@ Status MatMulBase::SetCostUnderStrategy(const mindspore::parallel::StrategyPtr &
std::shared_ptr<StrategyWithCost> swc =
std::make_shared<StrategyWithCost>(strategy, inputs_tensor_info_, outputs_tensor_info_);
swc->cost_list.push_back(result);
strategy_cost_.emplace_back(swc);
(void)strategy_cost_.emplace_back(swc);
return SUCCESS;
}

3
mindspore/ccsrc/frontend/parallel/ops_info/onehot_info.cc
View File

@ -53,7 +53,8 @@ Status OneHotInfo::CheckStrategy(const StrategyPtr &strategy) {
if (CheckStrategyValue(strategy, {outputs_shape_.at(0), inputs_shape_.at(1), inputs_shape_.at(2)}) != SUCCESS) {
return FAILED;
}
auto stra = strategy->GetInputDim().at(0);
auto strategies = strategy->GetInputDim();
auto stra = strategies.at(0);
bool invalid = false;
for (size_t i = 1; i < stra.size(); ++i) {
if (stra.at(i) != 1) {

16
mindspore/ccsrc/frontend/parallel/ops_info/operator_info.cc
View File

@ -394,7 +394,7 @@ void AddCommOpFusionType(const CNodePtr &comm_node, const AnfNodePtr &param_node
}
int32_t fusion_type = param_info->comm_fusion();
attrs[FUSION] = MakeValue<int64_t>(fusion_type);
prim->SetAttrs(attrs);
(void)prim->SetAttrs(attrs);
bool parallel_optimizer_comm_recompute = param_info->parallel_optimizer_comm_recompute();
std::string instance_name = prim->instance_name();
if (instance_name == PARALLEL_OPTIMIZER_ALLGATHER_NOT_COMPUTE && parallel_optimizer_comm_recompute &&
@ -412,7 +412,7 @@ void AddCommOpMeanFlag(const CNodePtr &comm_node) {
MS_EXCEPTION_IF_NULL(ParallelContext::GetInstance());
bool mean_flag = ParallelContext::GetInstance()->gradients_mean();
attrs[MEAN_FLAG] = MakeValue<bool>(mean_flag);
prim->SetAttrs(attrs);
(void)prim->SetAttrs(attrs);
}
void AddCommOpMirrorFlag(const CNodePtr &comm_node, bool do_mirror) {
@ -421,7 +421,7 @@ void AddCommOpMirrorFlag(const CNodePtr &comm_node, bool do_mirror) {
auto attrs = prim->attrs();
MS_EXCEPTION_IF_NULL(ParallelContext::GetInstance());
attrs[DO_MIRROR] = MakeValue<bool>(do_mirror);
prim->SetAttrs(attrs);
(void)prim->SetAttrs(attrs);
}
void AddCommOpAddAccuFlag(const CNodePtr &comm_node, bool add_accu) {
@ -430,7 +430,7 @@ void AddCommOpAddAccuFlag(const CNodePtr &comm_node, bool add_accu) {
auto attrs = prim->attrs();
MS_EXCEPTION_IF_NULL(ParallelContext::GetInstance());
attrs[ADD_ACCU] = MakeValue<bool>(add_accu);
prim->SetAttrs(attrs);
(void)prim->SetAttrs(attrs);
}
void AddCommOpParamFlag(const CNodePtr &comm_node) {
@ -1406,7 +1406,7 @@ Status GenerateStrategiesForIndependentInputs(int64_t stage_id, const Shapes &in
continue;
}
auto new_stra_arrays{stra_arrays};
new_stra_arrays[i][j] = new_stra_arrays[i][j] * dev_num_not_2_power;
new_stra_arrays[i][j] = new_stra_arrays[i][j] * UlongToLong(dev_num_not_2_power);
StrategyPtr new_stra = std::make_shared<Strategy>(stage_id, new_stra_arrays);
sp_vector->push_back(new_stra);
}
@ -1500,7 +1500,7 @@ Status OperatorInfo::SetCostUnderStrategyBase(const StrategyPtr &strategy) {
std::shared_ptr<StrategyWithCost> swc =
std::make_shared<StrategyWithCost>(strategy, inputs_tensor_info_, outputs_tensor_info_);
swc->cost_list.push_back(result);
strategy_cost_.emplace_back(swc);
(void)strategy_cost_.emplace_back(swc);
return SUCCESS;
}
@ -1617,7 +1617,6 @@ void OperatorInfo::ExactStrategiesAndRelatedEdges() {
ClearStrategyCost();
if (GenerateStrategies(0) != SUCCESS) {
MS_LOG(EXCEPTION) << "Strategy search for Operator " << name() << " failed.";
return;
}
SetIsStrategyCostExactTrue();
// re-init the previous edges
@ -1673,7 +1672,7 @@ int64_t OperatorInfo::ComputeOpAndPrevEdgeParameterInvolved() {
for (auto &p_edge : prev_edges) {
auto input_index = p_edge->next_op_input_index();
auto is_in_mem = p_edge->prev_operator()->operator_cost()->is_output_in_memory();
input_in_memory.emplace(std::make_pair(input_index, is_in_mem));
(void)input_in_memory.emplace(std::make_pair(input_index, is_in_mem));
}
operator_cost()->CalculateInputsInMemory(input_in_memory);
@ -1708,7 +1707,6 @@ Status OperatorInfo::CalculateMemoryCostForInference() {
// First, set the 'is_outputs_critical_' flag into OperatorCost.
if (is_output_critical_ == -1) {
MS_LOG(EXCEPTION) << "The critical flag is not set.";
return FAILED;
}
operator_cost()->set_output_critical(is_output_critical_);
// Set the memory cost in the 'strategy_cost_'

9
mindspore/ccsrc/frontend/parallel/ops_info/operator_info.h
View File

@ -54,7 +54,8 @@ class Edge;
class OperatorInfo {
public:
OperatorInfo(std::string name, Shapes inputs_shape, Shapes outputs_shape, PrimitiveAttrs attrs, OperatorCostPtr cost)
OperatorInfo(std::string name, Shapes inputs_shape, Shapes outputs_shape, PrimitiveAttrs attrs,
const OperatorCostPtr &cost)
: name_(std::move(name)),
inputs_shape_(std::move(inputs_shape)),
outputs_shape_(std::move(outputs_shape)),
@ -156,12 +157,12 @@ class OperatorInfo {
bool IsTmpIdentity() const;
void set_swc_index(int64_t, int64_t);
int64_t swc_index() { return swc_index_; }
int64_t swc_index() const { return swc_index_; }
// Approximate the list of available strategies
void ApproximateStrategies();
// Make the list of available strategies exact and re-init the related edges incident to this operator
void ExactStrategiesAndRelatedEdges();
bool is_strategy_cost_exact() { return is_strategy_cost_exact_; }
bool is_strategy_cost_exact() const { return is_strategy_cost_exact_; }
void SetIsStrategyCostExactTrue() { is_strategy_cost_exact_ = true; }
void ClearStrategyCost() { strategy_cost_.clear(); }
void CheckSelectedStrategy(const StrategyPtr &);
@ -245,7 +246,7 @@ class OperatorInfo {
float GetFloatAttr(const std::string &attr_name);
std::string GetStringAttr(const std::string &attr_name);
std::vector<int64_t> GetTupleIntAttr(const std::string &attr_name);
void ReportError(const std::string &error_msg) {
void ReportError(const std::string &error_msg) const {
if (is_auto_parallel_) {
MS_LOG(INFO) << error_msg;
} else {

8
mindspore/ccsrc/frontend/parallel/ops_info/random_choice_with_mask_info.cc
View File

@ -59,9 +59,9 @@ Status RandomChoiceWithMaskInfo::InferTensorMap() {
Shape output1_shape = outputs_shape_.at(1);
inputs_tensor_map_.clear();
inputs_tensor_map_.emplace_back(Shape(input0_shape.size(), -1));
outputs_tensor_map_.emplace_back(Shape(output0_shape.size(), -1));
outputs_tensor_map_.emplace_back(Shape(output1_shape.size(), -1));
(void)inputs_tensor_map_.emplace_back(Shape(input0_shape.size(), -1));
(void)outputs_tensor_map_.emplace_back(Shape(output0_shape.size(), -1));
(void)outputs_tensor_map_.emplace_back(Shape(output1_shape.size(), -1));
return SUCCESS;
}
@ -69,7 +69,7 @@ std::vector<StrategyPtr> RandomChoiceWithMaskInfo::GenerateOpStrategies(int64_t
Dimensions input_partitions(inputs_shape_[0].size(), 1);
Strategys strategies = {input_partitions};
std::vector<StrategyPtr> sp_vector;
sp_vector.emplace_back(std::make_shared<Strategy>(stage_id, strategies));
(void)sp_vector.emplace_back(std::make_shared<Strategy>(stage_id, strategies));
return sp_vector;
}

19
mindspore/ccsrc/frontend/parallel/ops_info/reduce_method_info.cc
View File

@ -145,7 +145,8 @@ bool IsDataParallelStrategy(const Dimensions &strategy, int32_t stage_id) {
}
Status ReduceMethod::InferForwardCommunication() {
Dimensions stra = strategy_->GetInputDim().at(0);
auto strategies = strategy_->GetInputDim();
Dimensions stra = strategies.at(0);
if (cross_batch_ && IsDataParallelStrategy(stra, stage_id_)) {
MS_LOG(INFO) << name_ << ": cross_batch is True, don't need to InferForwardCommunication";
return SUCCESS;
@ -225,7 +226,8 @@ ForwardOp CreateReduceMeanForwardOp(const std::vector<Group> &forward_group, con
}
Status ReduceMeanInfo::InferForwardCommunication() {
Dimensions stra = strategy_->GetInputDim().at(0);
auto strategies = strategy_->GetInputDim();
Dimensions stra = strategies.at(0);
if (cross_batch_ && IsDataParallelStrategy(stra, stage_id_)) {
MS_LOG(INFO) << name_ << ": cross_batch is True, don't need to InferForwardCommunication";
return SUCCESS;
@ -281,7 +283,7 @@ Status ReduceMeanInfo::InferForwardCommunication() {
return SUCCESS;
}
ForwardOp ReduceAnyInfo::CreateForwardOp(const std::vector<Group> &forward_group) {
ForwardOp ReduceAnyInfo::CreateForwardOp(const std::vector<Group> &forward_group) const {
// Create Cast to Int32 op
Operator op0 = CreateCastOp(kInt32);
@ -299,7 +301,8 @@ ForwardOp ReduceAnyInfo::CreateForwardOp(const std::vector<Group> &forward_group
}
Status ReduceAnyInfo::InferForwardCommunication() {
Dimensions stra = strategy_->GetInputDim().at(0);
auto strategies = strategy_->GetInputDim();
Dimensions stra = strategies.at(0);
if (cross_batch_ && IsDataParallelStrategy(stra, stage_id_)) {
MS_LOG(INFO) << name_ << ": cross_batch is True, don't need to InferForwardCommunication";
return SUCCESS;
@ -399,7 +402,8 @@ Status ArgMaxWithValueInfo::InferMirrorOps() {
Dimensions ReduceMethod::InferOutputStrategy() {
std::vector<int64_t> dim_list = reduce_dim();
Dimensions output_strategy;
Dimensions stra = strategy_->GetInputDim().at(0);
auto strategies = strategy_->GetInputDim();
Dimensions stra = strategies.at(0);
// if keepdims_ is true,then output strategy is same with input.
for (size_t i = 0; i < stra.size(); ++i) {
if (find(dim_list.begin(), dim_list.end(), SizeToLong(i)) != dim_list.end()) {
@ -786,7 +790,8 @@ Status SquareSumAllInfo::InferTensorInfo() {
}
Status SquareSumAllInfo::InferGroup() {
Dimensions stra = strategy_->GetInputDim().at(0);
auto strategies = strategy_->GetInputDim();
Dimensions stra = strategies.at(0);
forward_op_.clear();
std::vector<int64_t> dim_list = reduce_dim();
size_t size = stra.size();
@ -832,9 +837,11 @@ Status SquareSumAllInfo::ComputeReplaceGraph(const CNodePtr &cnode) {
GenerateGraph gen_g = GenerateGraph(attrs_);
if (gen_g.Init(cnode) != SUCCESS) {
MS_LOG(ERROR) << name_ << ": GenerateGraph Init failed";
return FAILED;
}
if (InferGroup() != SUCCESS) {
MS_LOG(ERROR) << name_ << ": Infer Group failed";
return FAILED;
}
MS_LOG(INFO) << name_ << ": The rank is " << g_device_manager->rank_index_in_stage();

4
mindspore/ccsrc/frontend/parallel/ops_info/reduce_method_info.h
View File

@ -33,7 +33,7 @@ namespace parallel {
class ReduceMethod : public OperatorInfo {
public:
ReduceMethod(const std::string &name, const Shapes &inputs_shape, const Shapes &outputs_shape,
const PrimitiveAttrs &attrs, OperatorCostPtr cost)
const PrimitiveAttrs &attrs, const OperatorCostPtr &cost)
: OperatorInfo(name, inputs_shape, outputs_shape, attrs, cost) {}
~ReduceMethod() override = default;
@ -130,7 +130,7 @@ class ReduceAnyInfo : public ReduceMethod {
protected:
Status InferForwardCommunication() override;
ForwardOp CreateForwardOp(const std::vector<Group> &forward_group);
ForwardOp CreateForwardOp(const std::vector<Group> &forward_group) const;
};
class ReduceMinInfo : public ReduceMethod {

20
mindspore/ccsrc/frontend/parallel/ops_info/roi_align_info.cc
View File

@ -30,7 +30,7 @@ Status ROIAlignInfo::GetAttrs() {
MS_LOG(ERROR) << name_ << ": Get primitive attr \"" << attr_key << "\" failed.";
return FAILED;
}
roi_align_attrs.emplace_back(std::make_pair(attr_key, attrs_[attr_key]));
(void)roi_align_attrs.emplace_back(std::make_pair(attr_key, attrs_[attr_key]));
}
return SUCCESS;
}
@ -61,8 +61,9 @@ Status ROIAlignInfo::CheckStrategy(const StrategyPtr &strategy) {
Status ROIAlignInfo::InferDevMatrixShape() {
dev_matrix_shape_.clear();
auto features_strategy = strategy_->GetInputDim().at(0);
auto rois_strategy = strategy_->GetInputDim().at(1);
auto strategies = strategy_->GetInputDim();
auto features_strategy = strategies.at(0);
auto rois_strategy = strategies.at(1);
dev_matrix_shape_ = {features_strategy[0], features_strategy[1], rois_strategy[0]};
return SUCCESS;
}
@ -85,9 +86,10 @@ Status ROIAlignInfo::InferBias() {
CheckGlobalDeviceManager();
int64_t rank = g_device_manager->rank_index_in_stage();
auto features_strategy = strategy_->GetInputDim().at(0);
auto strategies = strategy_->GetInputDim();
auto features_strategy = strategies.at(0);
auto features_shape = inputs_shape_.at(0);
auto rois_strategy = strategy_->GetInputDim().at(1);
auto rois_strategy = strategies.at(1);
auto rois_shape = inputs_shape_.at(1);
if (features_shape[0] % features_strategy[0] != 0 || rois_shape[0] % rois_strategy[0] != 0) {
return FAILED;
@ -202,7 +204,8 @@ Status ROIAlignInfo::InitForCostModel(const StrategyPtr &strategy, const Strateg
}
return FAILED;
}
auto features_strategy = strategy_->GetInputDim().at(0);
auto strategies = strategy_->GetInputDim();
auto features_strategy = strategies.at(0);
auto roi_align_cost = std::dynamic_pointer_cast<ROIAlignCost>(operator_cost());
auto pooled_height = GetValue<int64_t>(attrs_[POOLED_HEIGHT]);
auto pooled_width = GetValue<int64_t>(attrs_[POOLED_WIDTH]);
@ -213,14 +216,15 @@ Status ROIAlignInfo::InitForCostModel(const StrategyPtr &strategy, const Strateg
}
ReplaceGraphPtr ROIAlignInfo::replace_graph(const CNodePtr &cnode) {
auto features_strategy = strategy_->GetInputDim().at(0);
auto strategies = strategy_->GetInputDim();
auto features_strategy = strategies.at(0);
if (features_strategy[0] != 1 && ComputeReplaceGraph(cnode) != SUCCESS) {
MS_LOG(EXCEPTION) << name_ << ": ComputeReplaceGraph failed.";
}
return replace_graph_;
}
std::vector<int64_t> ROIAlignInfo::CreateRangeVector(int64_t upper_bound) {
std::vector<int64_t> ROIAlignInfo::CreateRangeVector(int64_t upper_bound) const {
std::vector<int64_t> range(upper_bound);
for (int64_t i = 0; i < upper_bound; ++i) {
range[LongToSize(i)] = i;

8
mindspore/ccsrc/frontend/parallel/ops_info/roi_align_info.h
View File

@ -50,11 +50,11 @@ class ROIAlignInfo : public OperatorInfo {
Status InferBias();
Status InferGroup();
Status ComputeReplaceGraph(const CNodePtr &cnode);
std::vector<int64_t> CreateRangeVector(int64_t upper_bound);
std::vector<int64_t> CreateRangeVector(int64_t upper_bound) const;
int64_t features_slice_size_;
int64_t rois_slice_size_;
int64_t bias_;
int64_t features_slice_size_ = 0;
int64_t rois_slice_size_ = 0;
int64_t bias_ = 0;
Group group_;
OperatorAttrs roi_align_attrs;
};

2
mindspore/ccsrc/frontend/parallel/ops_info/tensordot_info.cc
View File

@ -251,7 +251,7 @@ void TensorDotInfo::InferTensorMapAxesTuple(size_t size, const TensorMap &input_
if (!found) {
input_b_tensor_map.push_back(tmp_b_map_index.front());
tmp_b_map_index.erase(tmp_b_map_index.begin());
(void)tmp_b_map_index.erase(tmp_b_map_index.begin());
} else {
input_b_tensor_map.push_back(input_a_tensor_map[relevant_a_index]);
}

2
mindspore/ccsrc/frontend/parallel/ops_info/tensordot_info.h
View File

@ -45,6 +45,7 @@ class TensorDotInfo : public OperatorInfo {
~TensorDotInfo() override = default;
std::vector<StrategyPtr> GenerateOpStrategies(int64_t stage_id) override;
std::shared_ptr<Strategys> GenerateBatchStrategies() override;
Status SetCostUnderStrategy(const StrategyPtr &strategy) override;
Status PrepareStrategy(int32_t stage_id, size_t dev_num, Dimensions combined_partitions, size_t input0_shape_size,
size_t input1_shape_size, StrategyPtr *sp);
@ -55,7 +56,6 @@ class TensorDotInfo : public OperatorInfo {
Status InferDevMatrixShape() override;
Status InferTensorMap() override;
Status GetAttrs() override;
std::shared_ptr<Strategys> GenerateBatchStrategies() override;
void InferTensorMapAxesInt(const TensorMap &tensor_map_index);
void InferTensorMapAxesTuple(size_t size, const TensorMap &input_a_tensor_map, const TensorMap &tensor_map_index);
void ShowAxes();

12
mindspore/ccsrc/frontend/parallel/ops_info/unsorted_segment_op_info.cc
View File

@ -157,9 +157,9 @@ Status UnsortedSegmentOpInfo::InferTensorMap() {
return FAILED;
}
inputs_tensor_map_.emplace_back(std::move(tensor_map_in));
inputs_tensor_map_.emplace_back(std::move(tensor_map_in_index));
outputs_tensor_map_.emplace_back(std::move(tensor_map_out));
(void)inputs_tensor_map_.emplace_back(std::move(tensor_map_in));
(void)inputs_tensor_map_.emplace_back(std::move(tensor_map_in_index));
(void)outputs_tensor_map_.emplace_back(std::move(tensor_map_out));
return SUCCESS;
}
@ -242,7 +242,8 @@ std::shared_ptr<Strategys> UnsortedSegmentOpInfo::GenerateBatchStrategies() {
// a special case is when the shape input equals the shape of ids, we regard it as column slice,
// thus there is no need for repalce graphs
ReplaceGraphPtr UnsortedSegmentMinInfo::replace_graph(const CNodePtr &cnode) {
auto input_id_strategy = strategy_->GetInputDim().at(1);
auto strategies = strategy_->GetInputDim();
auto input_id_strategy = strategies.at(1);
// 1. the two input shapes are same, and the strategy is not all ones
if (std::any_of(input_id_strategy.begin(), input_id_strategy.end(), [](const int64_t &shard) { return shard > 1; })) {
if (ComputeReplaceGraph(cnode) != SUCCESS) {
@ -277,7 +278,8 @@ Status UnsortedSegmentMinInfo::ComputeReplaceGraph(const CNodePtr &cnode) {
// The UnsortedSegmentMaxInfo is almost same with UnsortedSegmentMinInfo
// Except the reduceMin op in the ComputeReplaceGraph is replaced with reduceMax op
ReplaceGraphPtr UnsortedSegmentMaxInfo::replace_graph(const CNodePtr &cnode) {
auto input_id_strategy = strategy_->GetInputDim().at(1);
auto strategies = strategy_->GetInputDim();
auto input_id_strategy = strategies.at(1);
// 1. the two input shapes are same, and the strategy is not all ones
if (std::any_of(input_id_strategy.begin(), input_id_strategy.end(), [](const int64_t &shard) { return shard > 1; })) {
if (ComputeReplaceGraph(cnode) != SUCCESS) {

11
mindspore/ccsrc/frontend/parallel/ops_info/unsorted_segment_op_info.h
View File

@ -41,7 +41,7 @@ constexpr size_t UNSORTEDSEGMENTOP_OUTPUTS_SIZE = 1;
class UnsortedSegmentOpInfo : public OperatorInfo {
public:
UnsortedSegmentOpInfo(const std::string &name, const Shapes &inputs_shape, const Shapes &outputs_shape,
const PrimitiveAttrs &attrs, OperatorCostPtr cost)
const PrimitiveAttrs &attrs, const OperatorCostPtr &cost)
: OperatorInfo(name, inputs_shape, outputs_shape, attrs, cost) {}
~UnsortedSegmentOpInfo() override = default;
@ -57,9 +57,6 @@ class UnsortedSegmentOpInfo : public OperatorInfo {
Status InferDevMatrixShape() override;
Status InferTensorMap() override;
Status GetAttrs() override;
private:
Status ComputeReplaceGraph(const CNodePtr &cnode);
};
class UnsortedSegmentSumInfo : public UnsortedSegmentOpInfo {
@ -92,11 +89,12 @@ class UnsortedSegmentMinInfo : public UnsortedSegmentOpInfo {
~UnsortedSegmentMinInfo() override = default;
ReplaceGraphPtr replace_graph(const CNodePtr &cnode) override;
Status InferForwardCommunication() override { return SUCCESS; }
protected:
Status ComputeReplaceGraph(const CNodePtr &cnode);
Status InferForwardCommunication() override { return SUCCESS; }
};
class UnsortedSegmentMaxInfo : public UnsortedSegmentOpInfo {
public:
UnsortedSegmentMaxInfo(const std::string &name, const Shapes &inputs_shape, const Shapes &outputs_shape,
@ -107,12 +105,11 @@ class UnsortedSegmentMaxInfo : public UnsortedSegmentOpInfo {
~UnsortedSegmentMaxInfo() override = default;
ReplaceGraphPtr replace_graph(const CNodePtr &cnode) override;
Status InferForwardCommunication() override { return SUCCESS; }
protected:
Status ComputeReplaceGraph(const CNodePtr &cnode);
Status InferForwardCommunication() override { return SUCCESS; }
};
} // namespace parallel
} // namespace mindspore
#endif // MINDSPORE_CCSRC_FRONTEND_PARALLEL_OPS_INFO_UNSORTEDSEGMENTOP_INFO_H_

1
mindspore/ccsrc/frontend/parallel/parallel_optimizer/opt_param_mgr.cc
View File

@ -148,7 +148,6 @@ class OptParamMgrImpl : public OptParamMgr {
return true;
}
private:
FuncGraphPtr root_;
int64_t DEFAULT_VAL = 64; // unit: KB
int64_t DIVISOR_K = 1024;

2
mindspore/ccsrc/frontend/parallel/parallel_stub/executor_manager_stub.h
View File

@ -27,7 +27,7 @@ class Executor;
class ExecutorManager {
public:
static ExecutorManager &Instance() {
static ExecutorManager instance;
static ExecutorManager instance = ExecutorManager();
return instance;
}
std::shared_ptr<Executor> GetExecutor(const std::string &device_name, uint32_t device_id);

5
mindspore/ccsrc/frontend/parallel/parallel_stub/executor_stub.h
View File

@ -27,8 +27,9 @@ class Executor {
public:
Executor(const std::string &device_name, uint32_t device_id) : device_name_(device_name), device_id_(device_id) {}
~Executor() = default;
bool CreateCommGroup(const std::string &group_name, std::vector<uint32_t> ranks) const { return true; }
bool DestroyCommGroup(const std::string &group_name) const { return true; }
bool CreateCommGroup(const std::string &, const std::vector<uint32_t> &) const { return true; }
bool DestroyCommGroup(const std::string &) const { return true; }
private:
std::string device_name_;

4
mindspore/ccsrc/frontend/parallel/tensor_layout/tensor_layout.cc
View File

@ -425,7 +425,9 @@ TensorLayout TensorLayout::TransferRepeatLayout() const {
Shape tensor_map(tensor_map_origin_.GetDimSize(), -1);
Shape tensor_shape(tensor_shape_origin_.array());
TensorLayout repeat;
repeat.InitFromVector(dev_mat, tensor_map, tensor_shape);
if (repeat.InitFromVector(dev_mat, tensor_map, tensor_shape) != SUCCESS) {
MS_LOG(EXCEPTION) << "Init from vector failed.";
}
return repeat;
}

8
mindspore/ccsrc/frontend/parallel/tensor_layout/tensor_layout.h
View File

@ -112,15 +112,15 @@ class TensorLayout {
void set_opt_weight_shard_step(int32_t step) { opt_weight_shard_step_ = step; }
int32_t opt_weight_shard_step() { return opt_weight_shard_step_; }
int32_t opt_weight_shard_step() const { return opt_weight_shard_step_; }
void set_opt_weight_shard_size(int32_t size) { opt_weight_shard_size_ = size; }
int32_t opt_weight_shard_size() { return opt_weight_shard_size_; }
int32_t opt_weight_shard_size() const { return opt_weight_shard_size_; }
void set_is_shared_param(bool is_shared_param) { is_shared_param_ = is_shared_param; }
bool is_shared_param() { return is_shared_param_; }
bool is_shared_param() const { return is_shared_param_; }
// Key for user data.
constexpr static char key[] = "TLayout";
@ -145,7 +145,7 @@ class TensorLayout {
bool skip_redistribution_ = false;
bool uniform_split_ = true;
bool layout_transfer_ = false;
int32_t field_size_ = 0;
int64_t field_size_ = 0;
Shape opt_shard_slice_shape_;
std::string opt_shard_group_ = ""; // for allgather
std::string opt_shard_mirror_group_ = ""; // for mirror ops

6
mindspore/ccsrc/frontend/parallel/tensor_layout/tensor_redistribution.cc
View File

@ -187,13 +187,13 @@ Status TensorRedistribution::InferRedistribution(const TensorLayout &from_layout
return Status::FAILED;
} else {
for (auto op : operator_infer.operator_vector()) {
operator_vector->insert(operator_vector->end(), op);
(void)operator_vector->insert(operator_vector->end(), op);
}
for (auto info : operator_infer.output_info_vector()) {
output_info_vector->insert(output_info_vector->end(), info);
(void)output_info_vector->insert(output_info_vector->end(), info);
}
for (auto opc : operator_infer.operator_list()) {
operator_list_.insert(operator_list_.end(), opc);
(void)operator_list_.insert(operator_list_.end(), opc);
}
}
return Status::SUCCESS;