From 66b20884dae75815f56dbaf99fc04c13085387a3 Mon Sep 17 00:00:00 2001 From: "wangwenhua1@huawei.com" Date: Fri, 13 Nov 2020 17:58:44 +0800 Subject: [PATCH] error message add --- ge/graph/common/transop_util.cc | 8 +++ ge/graph/common/transop_util.h | 2 + ge/graph/preprocess/graph_preprocess.cc | 64 +++++++++++++++++-- ge/graph/preprocess/insert_op/ge_aipp_op.cc | 6 ++ .../insert_op/util_insert_aipp_op.cc | 42 ++++++++---- ge/graph/preprocess/multi_batch_copy_graph.cc | 2 + ge/graph/preprocess/multi_batch_options.cc | 4 ++ 7 files changed, 112 insertions(+), 16 deletions(-) diff --git a/ge/graph/common/transop_util.cc b/ge/graph/common/transop_util.cc index 684ef3dc..f57f56a8 100755 --- a/ge/graph/common/transop_util.cc +++ b/ge/graph/common/transop_util.cc @@ -81,5 +81,13 @@ bool TransOpUtil::CheckPrecisionLoss(const ge::NodePtr &src_node) { return false; } return true; + + std::string TransOpUtil::TransopMapToString() { + std::string buffer; + for (auto it = transop_index_map_.begin(); it != transop_index_map_.end(); ++it) { + buffer += it->first + ","; + } + return buffer.substr(0, buffer.size() -1); + } } } // namespace ge diff --git a/ge/graph/common/transop_util.h b/ge/graph/common/transop_util.h index 8b10ad5c..3332e1fb 100644 --- a/ge/graph/common/transop_util.h +++ b/ge/graph/common/transop_util.h @@ -35,6 +35,8 @@ class GE_FUNC_HOST_VISIBILITY GE_FUNC_DEV_VISIBILITY TransOpUtil { static bool CheckPrecisionLoss(const NodePtr &src_node); + static std::string TransopMapToString(); + private: TransOpUtil(); diff --git a/ge/graph/preprocess/graph_preprocess.cc b/ge/graph/preprocess/graph_preprocess.cc index 98371426..7049fba1 100644 --- a/ge/graph/preprocess/graph_preprocess.cc +++ b/ge/graph/preprocess/graph_preprocess.cc @@ -128,6 +128,9 @@ static std::map output_type_str_to_datatype = { {"UINT32", ge::DT_UINT32}, {"UINT64", ge::DT_UINT64}, {"DOUBLE", ge::DT_DOUBLE}}; const char *const kMbatchSwitchnName = "mbatch-switch-name"; +const char *const kConstError1 = "Const is invalid scalar tensor."; +const char *const kConstError2 = "Const is invalid vector scalar."; +const char *const kConstError3 = "Const input data size is not equal with tensor desc shape"; // the size of user defined output datatype or format string after split by ":". const size_t kUserDefinedElementCount = 2; @@ -219,6 +222,9 @@ NodePtr CreateTransNode(const std::string &name, const std::string &node_type, c auto index = TransOpUtil::GetTransOpDataIndex(node_type); if (index < 0) { + ErrorManager::GetInstance().ATCReportErrMessage( + "E19025", {"situation", "reason"}, + {"The trans node type[" + node_type + "]", "it must be " + TransOpUtil::TransopMapToString()}); GELOGE(INTERNAL_ERROR, "The trans node type %s does not exists", node_type.c_str()); return nullptr; } @@ -387,6 +393,8 @@ Status RecoverTransRoadForVar(const NodePtr &var, const VarTransRoad &road) { auto trans_name = var->GetName() + "_trans_" + std::to_string(index++); auto ret = RecoverOneTransNodeForVar(trans_name, *iter, last_node, last_node); if (ret != SUCCESS) { + ErrorManager::GetInstance().ATCReportErrMessage( + "E15001", {"variable", "index", "type"}, {var->GetName(), std::to_string(index), iter->node_type}); GELOGE(INTERNAL_ERROR, "Failed to recover trans node for variable %s, index %d, type %s", var->GetName().c_str(), index, iter->node_type.c_str()); return INTERNAL_ERROR; @@ -419,6 +427,8 @@ Status RecoverTransRoadForVarRef(const std::set &nodes, const VarTransR auto trans_name = var->GetName() + "_trans_" + std::to_string(index++); auto ret = RecoverOneTransNodeForVarRef(trans_name, *iter, last_node, last_node); if (ret != SUCCESS) { + ErrorManager::GetInstance().ATCReportErrMessage( + "E15001", {"variable", "index", "type"}, {var->GetName(), std::to_string(index), iter->node_type}); GELOGE(INTERNAL_ERROR, "Failed to recover trans node for variable %s, index %d, type %s", var->GetName().c_str(), index, iter->node_type.c_str()); return INTERNAL_ERROR; @@ -571,6 +581,8 @@ Status CheckIfDynamicBatchScene(NodePtr &data_node, bool &is_dynamic_batch, Node std::string related_node_name; if (AttrUtils::GetStr(data_node->GetOpDesc(), kMbatchSwitchnName, related_node_name)) { if (related_node_name.empty()) { + ErrorManager::GetInstance().ATCReportErrMessage( + "E15002", {"opname", "value", "reason"}, {data_node->GetName(), "flag", "but the value is empty"}); GELOGE(INTERNAL_ERROR, "The data node %s has switchn node flag, but the value is empty", data_node->GetName().c_str()); return INTERNAL_ERROR; @@ -582,6 +594,9 @@ Status CheckIfDynamicBatchScene(NodePtr &data_node, bool &is_dynamic_batch, Node } } if (switchn_node == nullptr) { + ErrorManager::GetInstance().ATCReportErrMessage( + "E15002", {"opname", "value", "reason"}, + {data_node->GetName(), related_node_name, "but the value is empty"}); GELOGE(INTERNAL_ERROR, "The data node %s has switchn node %s, but can not find it on the graph", data_node->GetName().c_str(), related_node_name.c_str()); return INTERNAL_ERROR; @@ -682,6 +697,10 @@ Status ProcessInputNC1HWC0DynShape(NodePtr &node_ptr, bool &is_dynamic_batch, No ge::GeShape old_shape = input->GetShape(); bool support = ((old_format == FORMAT_NC1HWC0) || (old_format == FORMAT_NCHW) || (old_format == FORMAT_NHWC)); if (!support) { + ErrorManager::GetInstance().ATCReportErrMessage( + "E19014", {"opname", "value", "reason"}, + {op_desc->GetName(), "format[" + TypeUtils::FormatToSerialString(old_format) + "]", + "only support FORMAT_NC1HWC0,FORMAT_NCHW,FORMAT_NHWC"}); GELOGE(INTERNAL_ERROR, "The format [%s] is unsupported", TypeUtils::FormatToSerialString(old_format).c_str()); return FAILED; } @@ -762,6 +781,9 @@ Status GetStorageFormatAndShape(OpDescPtr &op_desc, const GeTensorDescPtr &tenso op_desc->GetName().c_str(), TypeUtils::FormatToSerialString(storage_format).c_str(), formats::JoinToString(storage_shape).c_str()); } else { + ErrorManager::GetInstance().ATCReportErrMessage( + "15003", {"opname", "format"}, + {op_desc->GetName(), TypeUtils::FormatToSerialString(storage_format)}); GELOGE(PARAM_INVALID, "Update node by storage format failed, storage_shape not set. " "node: [%s], storage_format [%s]", op_desc->GetName().c_str(), TypeUtils::FormatToSerialString(storage_format).c_str()); @@ -900,9 +922,14 @@ Status ProcessNetoutputNodeDynShape(NodePtr &node) { // check if is_output_adjust_hw_layout is set if (NeedUpdateFormatByOutputTypeParm(op_desc, index)) { if ((old_format != FORMAT_NCHW) && (old_format != FORMAT_NHWC) && (old_format != FORMAT_NC1HWC0)) { + ErrorManager::GetInstance().ATCReportErrMessage( + "E19014", {"opname", "value", "reason"}, + {op_desc->GetName(), "format[" + TypeUtils::FormatToSerialString(old_format) + "]", + "only support FORMAT_NC1HWC0,FORMAT_NCHW,FORMAT_NHWC"}); GELOGE(INTERNAL_ERROR, "Format is not one of NCHW, NHWC, NC1HWC0."); return FAILED; } + GeTensorDesc old_desc(old_shape, old_format, old_dtype); if (ProcessNetoutputNodeFp16Nc1hwc0DynShape(old_desc, net_output_input_desc, src_node) != SUCCESS) { GELOGE(INTERNAL_ERROR, "Process netoutput fp16 nc1hwc0."); @@ -1035,6 +1062,10 @@ Status GraphPrepare::CheckRefInputNode(const NodePtr &node, const std::string &i } bool is_acceptable = (acceptable_types.find(input_type) != acceptable_types.end()); if (!is_acceptable) { + ErrorManager::GetInstance().ATCReportErrMessage( + "E19014", {"opname", "value", "reason"}, + {op_desc->GetName(), "format[" + TypeUtils::FormatToSerialString(old_format) + "]", + "only support FORMAT_NC1HWC0,FORMAT_NCHW,FORMAT_NHWC"}); GELOGE(PARAM_INVALID, "The ref input of ref node %s[%s] must be ref node or variable, but %s[%s]isn't.", node->GetName().c_str(), node->GetType().c_str(), input_op_desc->GetName().c_str(), input_op_desc->GetType().c_str()); @@ -1127,6 +1158,9 @@ Status GraphPrepare::UpdateInput(const std::vector &user_input) { } if ((index < 0) || (static_cast(index) >= user_input.size())) { + std::string situation = "data op index[" + std::to_string(index) + "]"; + std::string reason = "it must less than user_input size[" + std::to_string(user_input.size()) + "]"; + ErrorManager::GetInstance().ATCReportErrMessage("E19025", {"situation", "reason"}, {situation, reason}); GELOGE(PARAM_INVALID, "user_input size = %zu, graph data op index = %ld.", user_input.size(), index); return FAILED; } @@ -1139,6 +1173,9 @@ Status GraphPrepare::UpdateInput(const std::vector &user_input) { if (need_check_internal_format) { bool is_internal = TypeUtils::IsInternalFormat(format) || TypeUtils::IsInternalFormat(origin_format); if (is_internal) { + ErrorManager::GetInstance().ATCReportErrMessage("E19025", {"situation", "reason"}, + {"Input format[" + TypeUtils::FormatToSerialString(format).c_str() + "] or origin_format[" + + TypeUtils::FormatToSerialString(origin_format).c_str() + "]", "it is not support"}); GELOGE(PARAM_INVALID, "Input format %s or origin_format %s is not support.", TypeUtils::FormatToSerialString(format).c_str(), TypeUtils::FormatToSerialString(origin_format).c_str()); @@ -1150,6 +1187,8 @@ Status GraphPrepare::UpdateInput(const std::vector &user_input) { uint32_t length = 1; bool type_ret = TypeUtils::GetDataTypeLength(data_type, length); if (!type_ret) { + ErrorManager::GetInstance().ATCReportErrMessage("E19025", {"situation", "reason"}, + {"Input datatype[" + TypeUtils::DataTypeToSerialString(data_type) + "]", "it is not support"}); GELOGE(PARAM_INVALID, "Input datatype %s is not support.", TypeUtils::DataTypeToSerialString(data_type).c_str()); return FAILED; @@ -1164,6 +1203,10 @@ Status GraphPrepare::UpdateInput(const std::vector &user_input) { return FAILED); bool size_check = (size != 0 && shape_size != size); if (size_check) { + std::string situation = "input data size[" + std::to_string(size) + + "] and shape_size[" + std::to_string(size) + "]"; + std::string reason = "because size != 0 and shape_size != size"; + ErrorManager::GetInstance().ATCReportErrMessage("E19025", {"situation", "reason"}, {situation, reason}); GELOGE(PARAM_INVALID, "input data size =%ld, shape_size =%ld.", size, shape_size); return FAILED; } @@ -1503,6 +1546,8 @@ Status GraphPrepare::VerifyConstOp(const NodePtr &node) { uint32_t length = 1; bool type_ret = TypeUtils::GetDataTypeLength(data_type, length); if (!type_ret) { + ErrorManager::GetInstance().ATCReportErrMessage("E19025", {"situation", "reason"}, + {"Input datatype[" + TypeUtils::DataTypeToSerialString(data_type) + "]", "it is not support"}); GELOGE(PARAM_INVALID, "Input datatype %s is not support.", TypeUtils::DataTypeToSerialString(data_type).c_str()); return FAILED; } @@ -1512,14 +1557,19 @@ Status GraphPrepare::VerifyConstOp(const NodePtr &node) { if (shape_size == 0) { if (ge_tensor_desc.GetShape().GetDims().size() == 0) { // shape = [], means it's a sclar tensor. - GE_CHK_BOOL_EXEC(data_size / length == 1, return PARAM_INVALID, "Const is invalid scalar tensor."); + GE_CHK_BOOL_EXEC(data_size / length == 1, + ErrorManager::GetInstance().ATCReportErrMessage("E10043", {"reason"}, {kConstError1}); + return PARAM_INVALID, kConstError1); } else { // shape = [x, y, 0,...], means it's a vector tensor that value is []. - GE_CHK_BOOL_EXEC(data_size == 0, return PARAM_INVALID, "Const is invalid vector scalar."); + GE_CHK_BOOL_EXEC(data_size == 0, + ErrorManager::GetInstance().ATCReportErrMessage("E10043", {"reason"}, {kConstError2}); + return PARAM_INVALID, kConstError2); } } else { - GE_CHK_BOOL_EXEC(data_size == static_cast(shape_size * length) && data_size != 0, return PARAM_INVALID, - "Const input data size is not equal with tensor desc shape"); + GE_CHK_BOOL_EXEC(data_size == static_cast(shape_size * length) && data_size != 0, + ErrorManager::GetInstance().ATCReportErrMessage("E10043", {"reason"}, {kConstError3}); + return PARAM_INVALID, kConstError3); } return SUCCESS; } @@ -1543,6 +1593,9 @@ Status GraphPrepare::CheckUserInput(const std::vector &user_input) { return GE_GRAPH_INIT_FAILED; } if ((index < 0) || (static_cast(index) >= user_input.size())) { + std::string situation = "data op index[" + std::to_string(index) + "]"; + std::string reason = "it must less than user_input size[" + std::to_string(user_input.size()) + "]"; + ErrorManager::GetInstance().ATCReportErrMessage("E19025", {"situation", "reason"}, {situation, reason}); GELOGE(GE_GRAPH_INIT_FAILED, "user_input size:%zu, data op index:%ld.", user_input.size(), index); return GE_GRAPH_INIT_FAILED; } @@ -1550,6 +1603,9 @@ Status GraphPrepare::CheckUserInput(const std::vector &user_input) { for (size_t i = 0; i < desc.GetShape().GetDimNum(); ++i) { if (desc.GetShape().GetDim(i) < 0) { + std::string situation = "data dim[" + std::to_string(i) + "][" + std::to_string(desc.GetShape().GetDim(i)) + "]" ; + std::string reason = "it need >= 0"; + ErrorManager::GetInstance().ATCReportErrMessage("E19025", {"situation", "reason"}, {situation, reason}); GELOGE(GE_GRAPH_INIT_FAILED, "data dim %zu is not supported, need >= 0, real:%ld.", i, desc.GetShape().GetDim(i)); return GE_GRAPH_INIT_FAILED; diff --git a/ge/graph/preprocess/insert_op/ge_aipp_op.cc b/ge/graph/preprocess/insert_op/ge_aipp_op.cc index 729c47de..474145af 100755 --- a/ge/graph/preprocess/insert_op/ge_aipp_op.cc +++ b/ge/graph/preprocess/insert_op/ge_aipp_op.cc @@ -790,16 +790,22 @@ Status AippOp::CreateAippData(const NodePtr &aipp_node) { int64_t batch_count = -1; if (GetDataDimN(data_node, ori_data_format, batch_count) != ge::SUCCESS) { + string errormsg = "Get data_node dims and transfer to nchw_dims failed!"; + ErrorManager::GetInstance().ATCReportErrMessage("E10043", {"reason"}, {errormsg}); GELOGE(PARAM_INVALID, "Get data_node dims and transfer to nchw_dims failed!"); return PARAM_INVALID; } if (batch_count <= 0) { + string errormsg = "Batch count[" + std::to_sting(batch_count) + "] is invalid, it must positive."; + ErrorManager::GetInstance().ATCReportErrMessage("E10043", {"reason"}, {errormsg}); GELOGE(PARAM_INVALID, "Batch count %ld is invalid", batch_count); return PARAM_INVALID; } int64_t max_dynamic_aipp_size = CalcMaxSize(batch_count); if (max_dynamic_aipp_size < 0) { + string errormsg = "The dynamic aipp size is not positive"; + ErrorManager::GetInstance().ATCReportErrMessage("E10043", {"reason"}, {errormsg}); GELOGE(PARAM_INVALID, "The dynamic aipp size is not positive."); return PARAM_INVALID; } diff --git a/ge/graph/preprocess/insert_op/util_insert_aipp_op.cc b/ge/graph/preprocess/insert_op/util_insert_aipp_op.cc index a1eb104d..8b33885b 100755 --- a/ge/graph/preprocess/insert_op/util_insert_aipp_op.cc +++ b/ge/graph/preprocess/insert_op/util_insert_aipp_op.cc @@ -37,6 +37,16 @@ using domi::AippOpParams; +#define AIPP_RETURN_STATUS_AND_REPROT_ERRORMSG(expr, _status, errormsg) \ + do { \ + bool b = (expr); \ + if (!b) { \ + GELOGE(_status, errormsg); \ + ErrorManager::GetInstance().ATCReportErrMessage("E10043", {"reason"}, {errormsg}); \ + return _status; \ + } \ + } while (0) + namespace ge { namespace { const char *const kMbatchSwitchnName = "mbatch-switch-name"; @@ -224,9 +234,10 @@ Status InsertNewOpUtil::CheckGraph(const ComputeGraphPtr &graph) { } } } - GE_CHK_BOOL_RET_STATUS((aippNodes.size() == 0) || (aippNodes.size() == next_nodes_cnt), PARAM_INVALID, - "Can not config part of outputs of Data node to support AIPP, config all " - "of the outputs of Data to support AIPP, or config none of them"); + AIPP_RETURN_STATUS_AND_REPROT_ERRORMSG((aippNodes.size() == 0) || (aippNodes.size() == next_nodes_cnt), + PARAM_INVALID, + "Can not config part of outputs of Data node to support AIPP, config all " + "of the outputs of Data to support AIPP, or config none of them"); std::unique_ptr aippParams(new (std::nothrow) domi::AippOpParams()); GE_CHECK_NOTNULL(aippParams); @@ -238,16 +249,19 @@ Status InsertNewOpUtil::CheckGraph(const ComputeGraphPtr &graph) { GE_CHK_STATUS(GetAippParams(currAippParam, aippNodes[i])); if (aippMode == domi::AippOpParams::static_) { - GE_CHK_BOOL_RET_STATUS(aippParams->input_format() == currAippParam->input_format(), PARAM_INVALID, - "The input_format of all aipp_ops after one Data should be the same"); - GE_CHK_BOOL_RET_STATUS(aippParams->src_image_size_w() == currAippParam->src_image_size_w(), PARAM_INVALID, - "The src_image_size_w of all aipp_ops after one Data should be the same"); - GE_CHK_BOOL_RET_STATUS(aippParams->src_image_size_h() == currAippParam->src_image_size_h(), PARAM_INVALID, - "The src_image_size_h of all aipp_ops after one Data should be the same"); + AIPP_RETURN_STATUS_AND_REPROT_ERRORMSG( + aippParams->input_format() == currAippParam->input_format(), + PARAM_INVALID, "The input_format of all aipp_ops after one Data should be the same"); + AIPP_RETURN_STATUS_AND_REPROT_ERRORMSG( + aippParams->src_image_size_w() == currAippParam->src_image_size_w(), + PARAM_INVALID, "The src_image_size_w of all aipp_ops after one Data should be the same"); + AIPP_RETURN_STATUS_AND_REPROT_ERRORMSG( + aippParams->src_image_size_h() == currAippParam->src_image_size_h(), + PARAM_INVALID, "The src_image_size_h of all aipp_ops after one Data should be the same"); } else { - GE_CHK_BOOL_RET_STATUS(aippParams->max_src_image_size() == currAippParam->max_src_image_size(), - PARAM_INVALID, - "The max_src_image_size of all aipp_ops after one Data should be the same"); + AIPP_RETURN_STATUS_AND_REPROT_ERRORMSG( + aippParams->max_src_image_size() == currAippParam->max_src_image_size(), + PARAM_INVALID, "The max_src_image_size of all aipp_ops after one Data should be the same"); } }); } @@ -290,6 +304,8 @@ Status InsertNewOpUtil::UpdateDataNodeByAipp(const ComputeGraphPtr &graph) { for (auto &switchn : updated_switchn) { auto data_iter = switchn_names_to_data.find(switchn->GetName()); if (data_iter == switchn_names_to_data.end()) { + string errormesg = "Failed to find relative data node by switchn[" + switchn->GetName() + "]"; + ErrorManager::GetInstance().ATCReportErrMessage("E10043", {"reason"}, {errormsg}); GELOGE(INTERNAL_ERROR, "Failed to find relative data node by switchn %s", switchn->GetName().c_str()); return INTERNAL_ERROR; } @@ -477,6 +493,8 @@ Status InsertNewOpUtil::UpdateDataBySwitchN(const NodePtr &switchn, const NodePt } } if (max_index >= switchn->GetOpDesc()->GetOutputsSize()) { + string errormesg = "No max size found from switchn node[" + switchn->GetName()+ "]"; + ErrorManager::GetInstance().ATCReportErrMessage("E10043", {"reason"}, {errormsg}); GELOGE(INTERNAL_ERROR, "No max size found from switchn node %s", switchn->GetName().c_str()); return INTERNAL_ERROR; } diff --git a/ge/graph/preprocess/multi_batch_copy_graph.cc b/ge/graph/preprocess/multi_batch_copy_graph.cc index c0ba89f4..8f42ad24 100644 --- a/ge/graph/preprocess/multi_batch_copy_graph.cc +++ b/ge/graph/preprocess/multi_batch_copy_graph.cc @@ -595,6 +595,8 @@ Status MultiBatchGraphCopyer::CheckCopyResult(const std::vector &start_ } auto dims = NodeUtils::GetOutputDesc(*node, kDataOutIndex).GetShape().GetDims(); if (!IsAllDimsPositive(dims)) { + ErrorManager::GetInstance().ATCReportErrMessage("E15004", {"opname", "shape"}, + {node->GetName(), formats::ShapeToString(dims)}); GELOGE(INTERNAL_ERROR, "Failed to copy multi batch graph, the node %s still has unknown shape %s", node->GetName().c_str(), formats::ShapeToString(dims).c_str()); return INTERNAL_ERROR; diff --git a/ge/graph/preprocess/multi_batch_options.cc b/ge/graph/preprocess/multi_batch_options.cc index 9909b0dc..cc0d2d5b 100644 --- a/ge/graph/preprocess/multi_batch_options.cc +++ b/ge/graph/preprocess/multi_batch_options.cc @@ -124,6 +124,8 @@ Status ParserDataToDynmaicInfo(const vector> &shapes, auto tmp_index = cur_data_index; for (size_t i = 0; i < static_cast(dynamic_dims_num); ++i) { if (tmp_index >= dynamic_gear_info.size()) { + ErrorManager::GetInstance().ATCReportErrMessage( + "E10045", {"name", "shape"}, {data_name, formats::JoinToString(data_shape)}); GELOGE(PARAM_INVALID, "Data: %s shape: %s make dynamic dims overflow", data_name.c_str(), formats::JoinToString(data_shape).c_str()); return FAILED; @@ -131,6 +133,8 @@ Status ParserDataToDynmaicInfo(const vector> &shapes, one_gear.push_back(dynamic_gear_info[tmp_index++]); } } else { + ErrorManager::GetInstance().ATCReportErrMessage( + "E10046", {"name", "shape"}, {data_name, formats::JoinToString(data_shape)}); GELOGE(PARAM_INVALID, "Dynamic dims num of data: %s shape: %s can not be more than one gear dynamic info size", data_name.c_str(), formats::JoinToString(data_shape).c_str()); return FAILED;