From 9a7cb76cb96b6c3eb2d3a4632039c3ee87dc86b0 Mon Sep 17 00:00:00 2001 From: liudingyan Date: Wed, 31 Mar 2021 10:43:49 +0800 Subject: [PATCH] modify geloge and add err report --- ge/hybrid/common/npu_memory_allocator.cc | 20 ++-- .../executor/hybrid_execution_context.cc | 5 +- .../executor/hybrid_model_async_executor.cc | 58 +++++------ .../hybrid_model_pipeline_executor.cc | 22 ++--- ge/hybrid/executor/hybrid_profiler.cc | 4 +- ge/hybrid/executor/node_state.cc | 14 +-- ge/hybrid/executor/rt_callback_manager.cc | 2 +- ge/hybrid/executor/subgraph_context.cc | 17 ++-- ge/hybrid/executor/subgraph_executor.cc | 25 +++-- ge/hybrid/executor/worker/execution_engine.cc | 4 +- .../executor/worker/shape_inference_engine.cc | 45 ++++----- ge/hybrid/model/graph_item.cc | 4 +- ge/hybrid/model/hybrid_model_builder.cc | 95 +++++++++++-------- 13 files changed, 172 insertions(+), 143 deletions(-) diff --git a/ge/hybrid/common/npu_memory_allocator.cc b/ge/hybrid/common/npu_memory_allocator.cc index 08f91f02..728a289b 100644 --- a/ge/hybrid/common/npu_memory_allocator.cc +++ b/ge/hybrid/common/npu_memory_allocator.cc @@ -38,9 +38,11 @@ AllocationAttr::AllocationAttr(void *try_reuse_addr) : AllocationAttr(0, try_reu NpuMemoryAllocator *NpuMemoryAllocator::GetAllocator() { int32_t device_id = 0; - if (rtGetDevice(&device_id) != RT_ERROR_NONE) { - GELOGE(RT_FAILED, "[Get][Device] Failed."); - REPORT_INNER_ERROR("E19999", "rtGetDevice failed when %s.", __FUNCTION__); + auto rt_result = rtGetDevice(&device_id); + if (rt_result != RT_ERROR_NONE) { + GELOGE(RT_FAILED, "[Get][Device] Failed, result:%d.", rt_result); + REPORT_INNER_ERROR("E19999", "rtGetDevice failed when NpuMemoryAllocator %s, result:%d.", + __FUNCTION__, rt_result); return nullptr; } @@ -58,9 +60,9 @@ void *NpuMemoryAllocator::Allocate(std::size_t size, AllocationAttr *attr) { } if (allocate_size == 0) { - GELOGE(MEMALLOC_FAILED, "[Check][Param:size_t]Memory size is 0, device_id = %u, size = %zu.", + GELOGE(MEMALLOC_FAILED, "[Check][Param:size_t]Memory size is 0, device_id = %u, size = %zu.", device_id_, allocate_size); - REPORT_INNER_ERROR("E19999", "Memory size is 0, device_id = %u, size = %zu when %s.", + REPORT_INNER_ERROR("E19999", "Memory size is 0, device_id = %u, size = %zu when %s.", device_id_, allocate_size, __FUNCTION__); return nullptr; } @@ -72,9 +74,9 @@ void *NpuMemoryAllocator::Allocate(std::size_t size, AllocationAttr *attr) { buffer = MemManager::Instance().HostMemInstance(RT_MEMORY_HBM).Malloc(allocate_size); } else { if (allocate_size > kMaxHbmMemorySize) { - GELOGE(PARAM_INVALID, "[Check][Param:size_t]Invalid HBM memory size: %zu bigger than limit:%lu, check invalid.", + GELOGE(PARAM_INVALID, "[Check][Param:size_t]Invalid HBM memory size: %zu bigger than limit:%lu, check invalid.", allocate_size, kMaxHbmMemorySize); - REPORT_CALL_ERROR("E19999", "Invalid HBM memory size: %zu bigger than limit:%lu, check invalid when %s.", + REPORT_CALL_ERROR("E19999", "Invalid HBM memory size: %zu bigger than limit:%lu, check invalid when %s.", allocate_size, kMaxHbmMemorySize, __FUNCTION__); return nullptr; } @@ -94,9 +96,9 @@ void *NpuMemoryAllocator::Allocate(std::size_t size, AllocationAttr *attr) { .Malloc(allocate_size, reinterpret_cast(try_reuse_addr), device_id_); } if (buffer == nullptr) { - GELOGE(MEMALLOC_FAILED, "[Malloc][Memory] Failed, device_id = %u, size = %zu.", + GELOGE(MEMALLOC_FAILED, "[Malloc][Memory] Failed, device_id = %u, size = %zu.", device_id_, allocate_size); - REPORT_CALL_ERROR("E19999", "malloc memory failed, device_id = %u, size = %zu when %s.", + REPORT_CALL_ERROR("E19999", "malloc memory failed, device_id = %u, size = %zu when %s.", device_id_, allocate_size, __FUNCTION__); return nullptr; } diff --git a/ge/hybrid/executor/hybrid_execution_context.cc b/ge/hybrid/executor/hybrid_execution_context.cc index dd964448..7cceadf3 100644 --- a/ge/hybrid/executor/hybrid_execution_context.cc +++ b/ge/hybrid/executor/hybrid_execution_context.cc @@ -59,9 +59,8 @@ Status GraphExecutionContext::Synchronize(rtStream_t rt_stream) { return SUCCESS; } - GELOGE(RT_FAILED, - "[Invoke][rtStreamSynchronize] failed, ret = %d", rt_ret); - REPORT_CALL_ERROR("E19999", + GELOGE(RT_FAILED, "[Invoke][rtStreamSynchronize] failed, ret = %d", rt_ret); + REPORT_CALL_ERROR("E19999", "invoke rtStreamSynchronize failed when GraphExecutionContext %s, ret = %d", __FUNCTION__, rt_ret); return RT_FAILED; } diff --git a/ge/hybrid/executor/hybrid_model_async_executor.cc b/ge/hybrid/executor/hybrid_model_async_executor.cc index 2720d72d..a97336d9 100644 --- a/ge/hybrid/executor/hybrid_model_async_executor.cc +++ b/ge/hybrid/executor/hybrid_model_async_executor.cc @@ -51,10 +51,10 @@ void HybridModelAsyncExecutor::SetModelName(const string &model_name) { } Status HybridModelAsyncExecutor::EnqueueData(const shared_ptr &data) { - if(data_inputer_->Push(data) != SUCCESS){ + if (data_inputer_->Push(data) != SUCCESS) { REPORT_CALL_ERROR("E19999", "Data queue is full, please call again later when %s, model_id %u.", __FUNCTION__, model_id_); - GELOGE(domi::DATA_QUEUE_ISFULL, + GELOGE(domi::DATA_QUEUE_ISFULL, "[Push][Data] Data queue is full, please call again later, model_id %u ", model_id_); return domi::DATA_QUEUE_ISFULL; } @@ -65,9 +65,9 @@ Status HybridModelAsyncExecutor::EnqueueData(const shared_ptr Status HybridModelAsyncExecutor::Start(const std::shared_ptr &listener) { GELOGD("HybridModelExecutor::Start IN, has listener = %d", listener != nullptr); std::lock_guard lk(mu_); - if(run_flag_){ - REPORT_INNER_ERROR("E19999", "Model already started when HybridModelAsyncExecutor %s, model_id:%u.", - __FUNCTION__, model_id_); + if (run_flag_) { + REPORT_INNER_ERROR("E19999", + "Model already started when HybridModelAsyncExecutor %s, model_id:%u.", __FUNCTION__, model_id_); GELOGE(INTERNAL_ERROR, "[Check][RunState] Model already started, model_id:%u.", model_id_); return INTERNAL_ERROR; } @@ -80,8 +80,8 @@ Status HybridModelAsyncExecutor::Start(const std::shared_ptr &lis return RunInternal(); }); - GE_CHK_BOOL_RET_STATUS(future_.valid(), INTERNAL_ERROR, - "[Check][RunState] Failed to start, model_id:%u.", model_id_); + GE_CHK_BOOL_RET_STATUS(future_.valid(), INTERNAL_ERROR, + "[Check][RunState] Failed to start, model_id:%u.", model_id_); GELOGD("HybridModelExecutor::Start successfully"); return SUCCESS; } @@ -115,8 +115,8 @@ Status HybridModelAsyncExecutor::Init() { executor_ = std::unique_ptr(new(std::nothrow) HybridModelExecutor(model_, device_id_, stream_)); GE_CHECK_NOTNULL(executor_); - GE_CHK_STATUS_RET(executor_->Init(), - "[Init][HybridModelExecutor] failed, model_id:%u.", model_id_); + GE_CHK_STATUS_RET(executor_->Init(), + "[Init][HybridModelExecutor] failed, model_id:%u.", model_id_); GE_CHK_STATUS_RET(DumpOpDebug(), "[Dump][OpDebug] failed, model_id:%u.", model_id_); GELOGI("HybridModel stage nums:%zu", model_->GetRootGraphItem()->NumGroups()); @@ -124,8 +124,8 @@ Status HybridModelAsyncExecutor::Init() { pipe_executor_ = std::unique_ptr(new(std::nothrow) HybridModelPipelineExecutor(model_, device_id_)); GE_CHECK_NOTNULL(pipe_executor_); - GE_CHK_STATUS_RET(pipe_executor_->Init(), - "[Init][HybridModelPipelineExecutor] failed, model_id:%u.", model_id_); + GE_CHK_STATUS_RET(pipe_executor_->Init(), + "[Init][HybridModelPipelineExecutor] failed, model_id:%u.", model_id_); } GE_CHK_STATUS_RET(InitInputDesc(), "[Init][InputDesc] failed, model_id:%u.", model_id_); @@ -136,8 +136,8 @@ Status HybridModelAsyncExecutor::Init() { Status HybridModelAsyncExecutor::PreRun(InputData ¤t_data, HybridModelExecutor::ExecuteArgs &args) { GE_CHK_STATUS_RET(SyncVarData(), "[Invoke][SyncVarData] failed, model_id:%u.", model_id_); RECORD_MODEL_EXECUTION_EVENT(executor_->GetContext(), "[SyncVarData] End"); - GE_CHK_STATUS_RET(PrepareInputs(current_data, args), - "[Invoke][PrepareInputs] failed to copy input data to model, model_id:%u.", model_id_); + GE_CHK_STATUS_RET(PrepareInputs(current_data, args), + "[Invoke][PrepareInputs] failed to copy input data to model, model_id:%u.", model_id_); RECORD_MODEL_EXECUTION_EVENT(executor_->GetContext(), "[CopyInputData] End"); return SUCCESS; } @@ -213,7 +213,7 @@ Status HybridModelAsyncExecutor::HandleResult(Status exec_ret, if (exec_ret != SUCCESS) { GELOGE(exec_ret, "[Check][Param:Status] failed to execute graph. model_id = %u", model_id_); - REPORT_INNER_ERROR("E19999", + REPORT_INNER_ERROR("E19999", "failed to execute graph when HybridModelAsyncExecutor %s. model_id = %u", __FUNCTION__, model_id_); return OnComputeDone(data_id, INTERNAL_ERROR, output_tensor_info_list); } @@ -250,10 +250,10 @@ Status HybridModelAsyncExecutor::SyncVarData() { Status HybridModelAsyncExecutor::PrepareInputs(const InputData ¤t_data, HybridModelExecutor::ExecuteArgs &args) { if (current_data.blobs.size() < input_tensor_desc_.size()) { - GELOGE(PARAM_INVALID, + GELOGE(PARAM_INVALID, "[Check][Size]Blob size mismatches, expect at least %zu, but got %zu, model_id = %u", input_tensor_desc_.size(), current_data.blobs.size(), model_id_); - REPORT_INNER_ERROR("E19999", + REPORT_INNER_ERROR("E19999", "Blob size mismatches, expect at least %zu, but got %zu when HybridModelAsyncExecutor %s, model_id = %u.", input_tensor_desc_.size(), current_data.blobs.size(), __FUNCTION__, model_id_); return PARAM_INVALID; @@ -267,12 +267,12 @@ Status HybridModelAsyncExecutor::PrepareInputs(const InputData ¤t_data, Hy auto tensor_size = input_sizes_[input_index]; if (is_input_dynamic_[input_index]) { if (input_index >= current_data.shapes.size()) { - GELOGE(PARAM_INVALID, + GELOGE(PARAM_INVALID, "[Check][Range]Shape index out of range, index = %zu, shape size = %zu model_id = %u.", input_index, current_data.shapes.size(), model_id_); - REPORT_INNER_ERROR("E19999", + REPORT_INNER_ERROR("E19999", "Shape index out of range, index = %zu, shape size = %zu when HybridModelAsyncExecutor %s, model_id = %u.", - input_index, current_data.shapes.size(), __FUNCTION__, model_id_); + input_index, current_data.shapes.size(), __FUNCTION__, model_id_); return PARAM_INVALID; } auto &tensor_desc = input_tensor_desc_[input_index]; @@ -287,12 +287,12 @@ Status HybridModelAsyncExecutor::PrepareInputs(const InputData ¤t_data, Hy } // range[k].second can be -1 if (shape.GetDim(k) < range[k].first || (range[k].second >= 0 && shape.GetDim(k) > range[k].second)) { - GELOGE(PARAM_INVALID, + GELOGE(PARAM_INVALID, "[Check][Range]Dim out of range, shape idx = %zu, dim idx = %zu, dim = %ld, range = [%ld, %ld], model_id = %u.", input_index, k, shape.GetDim(k), range[k].first, range[k].second, model_id_); - REPORT_INNER_ERROR("E19999", + REPORT_INNER_ERROR("E19999", "Dim out of range, shape idx = %zu, dim idx = %zu, dim = %ld, range = [%ld, %ld], model_id = %u.", - input_index, k, shape.GetDim(k), range[k].first, range[k].second, model_id_); + input_index, k, shape.GetDim(k), range[k].first, range[k].second, model_id_); return PARAM_INVALID; } } @@ -318,14 +318,14 @@ Status HybridModelAsyncExecutor::PrepareInputs(const InputData ¤t_data, Hy const DataBuffer &data_buf = blobs[input_index]; auto mem_size = static_cast(tensor_size); - if(mem_size < data_buf.length){ - REPORT_INNER_ERROR("E19999", + if (mem_size < data_buf.length) { + REPORT_INNER_ERROR("E19999", "input data size(%lu) does not match model required size(%lu) when %s, ret failed, model_id = %u.", data_buf.length, mem_size, __FUNCTION__, model_id_); GELOGE(PARAM_INVALID, "[Check][Size]input data size(%lu) does not match model required size(%lu), ret failed, model_id = %u.", data_buf.length, mem_size, model_id_); - return PARAM_INVALID; + return PARAM_INVALID; } if (data_buf.length > 0) { GELOGI("[IMAS]CopyPlainData memcpy graph_%u type[F] output[%zu] memaddr[%p] mem_size[%zu] datasize[%lu]", @@ -433,7 +433,7 @@ Status HybridModelAsyncExecutor::CopyOutputs(HybridModelExecutor::ExecuteArgs &a GELOGE(INTERNAL_ERROR, "[Check][Size]output[%zu] tensor size(%zu) is not enough for output shape [%s], model_id = %u.", i, output_tensor.GetSize(), tensor_desc->GetShape().ToString().c_str(), model_id_); - REPORT_INNER_ERROR("E19999", + REPORT_INNER_ERROR("E19999", "output[%zu] tensor size(%zu) is not enough for output shape [%s] model_id = %u," " when HybridModelAsyncExecutor %s.", i, output_tensor.GetSize(), tensor_desc->GetShape().ToString().c_str(), model_id_, __FUNCTION__); @@ -513,14 +513,14 @@ Status HybridModelAsyncExecutor::Execute(const vector &inputs, vector< } HybridModelExecutor::ExecuteArgs args; - GE_CHK_STATUS_RET(PrepareInputs(input_data, args), + GE_CHK_STATUS_RET(PrepareInputs(input_data, args), "[Invoke][PrepareInputs]Failed to copy input data to model, model_id = %u", model_id_); GELOGD("Done copying input data successfully."); GE_CHK_STATUS_RET(executor_->Execute(args), "[Invoke][Execute] Failed, model_id = %u.", model_id_); std::vector output_tensor_info_list; OutputData output_data; - GE_CHK_STATUS_RET(CopyOutputs(args, &output_data, output_tensor_info_list), + GE_CHK_STATUS_RET(CopyOutputs(args, &output_data, output_tensor_info_list), "[Invoke][CopyOutputs]Failed to copy outputs, model_id = %u.", model_id_); GELOGD("Done copying output data successfully. output count = %zu", output_tensor_info_list.size()); @@ -572,7 +572,7 @@ Status HybridModelAsyncExecutor::DumpOpDebug() { loop_cond = const_cast(varible_loop_cond->GetData()); } data_dumper_.SetLoopAddr(global_step, loop_per_iter, loop_cond); - GE_CHK_STATUS_RET(data_dumper_.LoadDumpInfo(), + GE_CHK_STATUS_RET(data_dumper_.LoadDumpInfo(), "[Invoke][LoadDumpInfo] failed in hybrid engine, model_id = %u.", model_id_); GELOGD("Dump op debug SUCCESS in hybrid engine"); } diff --git a/ge/hybrid/executor/hybrid_model_pipeline_executor.cc b/ge/hybrid/executor/hybrid_model_pipeline_executor.cc index 098c194c..0633eeb5 100644 --- a/ge/hybrid/executor/hybrid_model_pipeline_executor.cc +++ b/ge/hybrid/executor/hybrid_model_pipeline_executor.cc @@ -59,9 +59,9 @@ Status StageExecutor::Start(const std::vector &inputs, const std::v task_queue_.Pop(task_info); GELOGD("[Executor: %d] Got task, stage = %d, iteration = %ld", id_, task_info.stage, task_info.iteration); if (task_info.iteration >= pipe_config_->iteration_end) { - GELOGE(INTERNAL_ERROR, "[Check][Range][Executor: %d] Unexpected iteration: %d.", + GELOGE(INTERNAL_ERROR, "[Check][Range][Executor: %d] Unexpected iteration: %ld.", id_, task_info.iteration); - REPORT_INNER_ERROR("E19999", "[Executor: %d] Unexpected iteration: %d when StageExecutor %s.", + REPORT_INNER_ERROR("E19999", "[Executor: %d] Unexpected iteration: %ld when StageExecutor %s.", id_, task_info.iteration, __FUNCTION__); return INTERNAL_ERROR; } @@ -69,16 +69,16 @@ Status StageExecutor::Start(const std::vector &inputs, const std::v if (task_info.event != nullptr) { GELOGD("[%d] Add StreamWaitEvent", id_); GE_CHK_RT_RET(rtStreamWaitEvent(stream_, task_info.event)); - RECORD_MODEL_EXECUTION_EVENT(&context_, "[iteration = %d] [Stage = %d] End", task_info.iteration - 1, + RECORD_MODEL_EXECUTION_EVENT(&context_, "[iteration = %ld] [Stage = %d] End", task_info.iteration - 1, task_info.stage); } - RECORD_MODEL_EXECUTION_EVENT(&context_, "[iteration = %d] [Stage = %d] Start", task_info.iteration, + RECORD_MODEL_EXECUTION_EVENT(&context_, "[iteration = %lld] [Stage = %d] Start", task_info.iteration, task_info.stage); if (task_info.stage == 0) { GELOGD("[Executor: %d] To ResetExecutionContext", id_); - GE_CHK_STATUS_RET(ResetExecutionContext(context_), + GE_CHK_STATUS_RET(ResetExecutionContext(context_), "[Invoke][ResetExecutionContext][Executor: %d] Failed to reset context", id_); context_.iteration = task_info.iteration; GE_CHK_STATUS_RET_NOLOG(SetInputs(inputs, input_desc)); @@ -96,10 +96,10 @@ Status StageExecutor::Start(const std::vector &inputs, const std::v auto sync_result = Synchronize(); if (sync_result != SUCCESS) { - GELOGE(sync_result, - "[Invoke][Synchronize][Executor: %d] Failed to sync result:%d. iteration = %d", + GELOGE(sync_result, + "[Invoke][Synchronize][Executor: %d] Failed to sync result:%d. iteration = %ld", id_, sync_result, task_info.iteration); - REPORT_CALL_ERROR("E19999", "[Executor: %d] Failed to sync result:%d when StageExecutor %s. iteration = %d", + REPORT_CALL_ERROR("E19999", "[Executor: %d] Failed to sync result:%d when StageExecutor %s. iteration = %ld", id_, sync_result, __FUNCTION__, task_info.iteration); context_.profiler->Dump(std::cout); context_.callback_manager->Destroy(); @@ -107,11 +107,11 @@ Status StageExecutor::Start(const std::vector &inputs, const std::v return sync_result; } - RECORD_MODEL_EXECUTION_EVENT(&context_, "[iteration = %d] [Stage = %d] End", task_info.iteration, task_info.stage); + RECORD_MODEL_EXECUTION_EVENT(&context_, "[iteration = %ld] [Stage = %d] End", task_info.iteration, task_info.stage); // if not end stage if (task_info.stage >= pipe_config_->num_stages - 1) { - RECORD_MODEL_EXECUTION_EVENT(&context_, "[iteration = %d] Schedule End", task_info.iteration); + RECORD_MODEL_EXECUTION_EVENT(&context_, "[iteration = %ld] Schedule End", task_info.iteration); GELOGD("[Executor: %d] End of iteration [%ld]", id_, task_info.iteration); context_.callback_manager->Destroy(); RuntimeInferenceContext::DestroyContext(std::to_string(context_.context_id)); @@ -261,7 +261,7 @@ Status HybridModelPipelineExecutor::Execute(HybridModelExecutor::ExecuteArgs &ar if (ret != SUCCESS) { GELOGE(ret, "[Invoke][Synchronize] failed for [Executor: %zu].", i); REPORT_CALL_ERROR("E19999", "[Executor: %zu] failed to Synchronize result when HybridModelPipelineExecutor %s.", - i, __FUNCTION__); + i, __FUNCTION__); has_error = true; continue; } diff --git a/ge/hybrid/executor/hybrid_profiler.cc b/ge/hybrid/executor/hybrid_profiler.cc index bbe82dad..de2322b9 100644 --- a/ge/hybrid/executor/hybrid_profiler.cc +++ b/ge/hybrid/executor/hybrid_profiler.cc @@ -49,9 +49,9 @@ void HybridProfiler::RecordEvent(EventType event_type, const char *fmt, ...) { va_end(args); auto index = counter_++; if (index >= static_cast(events_.size())) { - GELOGE(INTERNAL_ERROR, + GELOGE(INTERNAL_ERROR, "[Check][Range]index out of range. index = %d, max event size = %zu", index, events_.size()); - REPORT_INNER_ERROR("E19999", "index out of range when HybridProfiler %s. index = %d, max event size = %zu", + REPORT_INNER_ERROR("E19999", "index out of range when HybridProfiler %s. index = %d, max event size = %zu", __FUNCTION__, index, events_.size()); return; } diff --git a/ge/hybrid/executor/node_state.cc b/ge/hybrid/executor/node_state.cc index f30901c9..ddeeaae7 100644 --- a/ge/hybrid/executor/node_state.cc +++ b/ge/hybrid/executor/node_state.cc @@ -67,10 +67,10 @@ Status ShapeInferenceState::UpdateInputShape(int idx, const GeTensorDesc &target Format format = input_desc.GetFormat(); DataType data_type = input_desc.GetDataType(); if (TensorUtils::CalcTensorMemSize(shape, format, data_type, tensor_size) != GRAPH_SUCCESS) { - GELOGE(FAILED, "[Invoke][CalcTensorMemSize] failed for [%s].", + GELOGE(FAILED, "[Invoke][CalcTensorMemSize] failed for [%s].", node_item.NodeName().c_str()); - REPORT_CALL_ERROR("E19999", "CalcTensorMemSize failed for [%s] when ShapeInferenceState %s.", - node_item.NodeName().c_str(), __FUNCTION__); + REPORT_CALL_ERROR("E19999", "CalcTensorMemSize failed for [%s] when ShapeInferenceState %s.", + node_item.NodeName().c_str(), __FUNCTION__); return FAILED; } } @@ -124,18 +124,18 @@ Status ShapeInferenceState::AwaitShapesReady(const GraphExecutionContext &contex } if (context.GetStatus() != SUCCESS) { - GELOGE(FAILED, "[Check][Status][%s] Await pending shape cancelled.", + GELOGE(FAILED, "[Check][Status][%s] Await pending shape cancelled.", node_item.NodeName().c_str()); - REPORT_CALL_ERROR("E19999", "[%s] Await pending shape cancelled when %s.", + REPORT_CALL_ERROR("E19999", "[%s] Await pending shape cancelled when %s.", node_item.NodeName().c_str(), __FUNCTION__); break; } } if (!wait_success) { - GELOGE(FAILED, "[Check][Status][%s] Wait for shape timeout:%d.", + GELOGE(FAILED, "[Check][Status][%s] Wait for shape timeout:%d.", node_item.NodeName().c_str(), kWaitInternal); - REPORT_CALL_ERROR("E19999", "[%s] Wait for shape timeout:%d when %s.", + REPORT_CALL_ERROR("E19999", "[%s] Wait for shape timeout:%d when %s.", node_item.NodeName().c_str(), kWaitInternal, __FUNCTION__); return FAILED; } diff --git a/ge/hybrid/executor/rt_callback_manager.cc b/ge/hybrid/executor/rt_callback_manager.cc index 9ceba0a1..51a93938 100644 --- a/ge/hybrid/executor/rt_callback_manager.cc +++ b/ge/hybrid/executor/rt_callback_manager.cc @@ -76,7 +76,7 @@ Status CallbackManager::CallbackProcess(rtContext_t context) { auto rt_err = rtEventSynchronize(event); if (rt_err != RT_ERROR_NONE) { GELOGE(RT_FAILED, "[Invoke][rtEventSynchronize] failed. ret = %d", rt_err); - REPORT_CALL_ERROR("E19999", + REPORT_CALL_ERROR("E19999", "Invoke rtEventSynchronize failed when CallbackManager %s, ret = %d.", __FUNCTION__, rt_err); GE_CHK_RT(rtEventDestroy(event)); return RT_FAILED; diff --git a/ge/hybrid/executor/subgraph_context.cc b/ge/hybrid/executor/subgraph_context.cc index 34cc2bbf..f198de8b 100644 --- a/ge/hybrid/executor/subgraph_context.cc +++ b/ge/hybrid/executor/subgraph_context.cc @@ -50,10 +50,10 @@ NodeStatePtr SubgraphContext::GetOrCreateNodeState(const NodeItem *node_item) { Status SubgraphContext::SetInput(int index, const TensorValue &tensor) { if (static_cast(index) >= all_inputs_.size()) { GELOGE(INTERNAL_ERROR, - "[Check][Param:index]output index output range. all input num = %zu, input index = %d", + "[Check][Param:index]input index out of range. all input num = %zu, input index = %d", all_inputs_.size(), index); - REPORT_INNER_ERROR("E19999", - "input param index out range when SubgraphContext %s, all input num = %zu, input index = %d.", + REPORT_INNER_ERROR("E19999", + "input param index out of range when SubgraphContext %s, all input num = %zu, input index = %d.", __FUNCTION__, all_inputs_.size(), index); return INTERNAL_ERROR; } @@ -70,9 +70,9 @@ Status SubgraphContext::SetOutput(const NodeItem &node_item, int output_index, c auto index = node_item.output_start + output_index; if ((output_index >= node_item.num_outputs) || (static_cast(index) >= all_outputs_.size())) { GELOGE(INTERNAL_ERROR, - "[Check][Param:output_index]output index output range. all output num = %zu, node_item = %s," + "[Check][Param:output_index]output index out of range. all output num = %zu, node_item = %s," "output index = %d.", all_outputs_.size(), node_item.DebugString().c_str(), output_index); - REPORT_INNER_ERROR("E19999", "output index output range when SubgraphContext %s. " + REPORT_INNER_ERROR("E19999", "output index out of range when SubgraphContext %s. " "all output num = %zu, node_item = %s, output index = %d.", __FUNCTION__, all_outputs_.size(), node_item.DebugString().c_str(), output_index); return INTERNAL_ERROR; @@ -129,9 +129,10 @@ Status SubgraphContext::Await(const NodePtr &node) { void SubgraphContext::OnError(Status error) { if (error != END_OF_SEQUENCE) { - GELOGE(error, "[Check][Param:error][%s] Error occurred while executing graph.", graph_item_->GetName().c_str()); - REPORT_INNER_ERROR("E19999", "[%s] Error occurred while executing graph when SubgraphContext %s.", - graph_item_->GetName().c_str(), __FUNCTION__); + GELOGE(error, "[Check][Param:error][%s] Error:%d occurred while executing graph.", + graph_item_->GetName().c_str(), error); + REPORT_INNER_ERROR("E19999", "[%s] Error:%d occurred while executing graph when SubgraphContext %s.", + graph_item_->GetName().c_str(), error, __FUNCTION__); } node_done_manager_.Destroy(); } diff --git a/ge/hybrid/executor/subgraph_executor.cc b/ge/hybrid/executor/subgraph_executor.cc index 3bd78579..4c8b1bc1 100644 --- a/ge/hybrid/executor/subgraph_executor.cc +++ b/ge/hybrid/executor/subgraph_executor.cc @@ -44,7 +44,7 @@ Status SubgraphExecutor::Init(const std::vector &inputs, const std::vector &input_desc) { subgraph_context_.reset(new(std::nothrow)SubgraphContext(graph_item_, context_)); GE_CHECK_NOTNULL(subgraph_context_); - GE_CHK_STATUS_RET(subgraph_context_->Init(), + GE_CHK_STATUS_RET(subgraph_context_->Init(), "[Init][SubgraphContext][%s] Failed to init subgraph context.", graph_item_->GetName().c_str()); shape_inference_engine_.reset(new(std::nothrow) ShapeInferenceEngine(context_, subgraph_context_.get())); @@ -68,11 +68,12 @@ Status SubgraphExecutor::InitInputsForUnknownShape(const std::vectorGetInputNodes(); if (inputs.size() < input_nodes.size()) { - GELOGE(INTERNAL_ERROR, + GELOGE(INTERNAL_ERROR, "[Check][Size][%s] Number of inputs [%zu] is not sufficient for subgraph which needs [%zu] inputs.", graph_item_->GetName().c_str(), inputs.size(), input_nodes.size()); - REPORT_INNER_ERROR("E19999", - "Number of inputs [%zu] is not sufficient for subgraph which needs [%zu] inputs when SubgraphExecutor %s.", + REPORT_INNER_ERROR("E19999", + "[%s] Number of inputs [%zu] is not sufficient for subgraph which needs [%zu] inputs," + "check invalid when SubgraphExecutor %s.", graph_item_->GetName().c_str(), inputs.size(), input_nodes.size(), __FUNCTION__); return INTERNAL_ERROR; } @@ -117,8 +118,9 @@ Status SubgraphExecutor::InitInputsForKnownShape(const std::vector GELOGE(INTERNAL_ERROR, "[Check][Size][%s] Number of inputs [%zu] is not sufficient for subgraph which needs at lease [%d] inputs", graph_item_->GetName().c_str(), inputs.size(), parent_input_index + 1); - REPORT_INNER_ERROR("E19999", - "[%s] Number of inputs [%zu] is not sufficient for subgraph which needs at lease [%d] inputs when %s.", + REPORT_INNER_ERROR("E19999", + "[%s] Number of inputs [%zu] is not sufficient for subgraph which needs at lease [%d] inputs," + "check invalid when %s.", graph_item_->GetName().c_str(), inputs.size(), parent_input_index + 1, __FUNCTION__); return INTERNAL_ERROR; } @@ -387,9 +389,10 @@ Status SubgraphExecutor::GetOutputs(vector &outputs, std::vectorGetName().c_str()); if (outputs.size() != output_desc.size()) { GELOGE(INTERNAL_ERROR, - "[Check][Size]Number of output tensors(%zu) mismatch number of output tensor desc(%zu).", + "[Check][Size]Number of outputs(%zu) mismatch number of output_desc(%zu).", outputs.size(), output_desc.size()); - REPORT_INNER_ERROR("E19999", "Number of output tensors(%zu) mismatch number of output tensor desc(%zu) when %s.", + REPORT_INNER_ERROR("E19999", "Number of outputs(%zu) mismatch number of output_desc(%zu)," + "check invlid when SubgraphExecutor %s.", outputs.size(), output_desc.size(), __FUNCTION__); return INTERNAL_ERROR; } @@ -413,9 +416,11 @@ Status SubgraphExecutor::SetOutputsToParentNode(TaskContext &task_context) { "[Invoke][GetOutputDescList][%s] Failed to get output tensor desc.", graph_item_->GetName().c_str()); if (outputs.size() != output_desc_list.size()) { - GELOGE(INTERNAL_ERROR, "[Check][Size][%s] num output tensors = %zu, num output tensor desc = %zu", + GELOGE(INTERNAL_ERROR, "[Check][Size][%s] num of output tensors = %zu, num of output tensor desc = %zu not equal", graph_item_->GetName().c_str(), outputs.size(), output_desc_list.size()); - REPORT_INNER_ERROR("E19999", "[%s] num output tensors = %zu, num output tensor desc = %zu when %s", + REPORT_INNER_ERROR("E19999", + "%s num of output tensors = %zu, num of output tensor desc = %zu not equal," + "check invalid when SubgraphExecutor %s", graph_item_->GetName().c_str(), outputs.size(), output_desc_list.size(), __FUNCTION__); return INTERNAL_ERROR; } diff --git a/ge/hybrid/executor/worker/execution_engine.cc b/ge/hybrid/executor/worker/execution_engine.cc index 797c0092..8dfdb476 100755 --- a/ge/hybrid/executor/worker/execution_engine.cc +++ b/ge/hybrid/executor/worker/execution_engine.cc @@ -105,7 +105,7 @@ Status NodeDoneCallback::PrepareConstInputs(const NodeItem &node_item) { "[Check][Size][%s] Tensor size is not enough. output index = %d, required size = %ld, tensor = %s.", node_item.NodeName().c_str(), output_idx, tensor_size, output_tensor->DebugString().c_str()); - REPORT_INNER_ERROR("E19999", + REPORT_INNER_ERROR("E19999", "[%s] Tensor size is not enough. output index = %d, required size = %ld, tensor = %s when %s.", node_item.NodeName().c_str(), output_idx, tensor_size, output_tensor->DebugString().c_str(), __FUNCTION__); @@ -423,7 +423,7 @@ Status ExecutionEngine::ValidateInputTensors(const NodeState &node_state, const "[Check][Size] for [%s] Input[%d]: tensor size mismatches. expected: %ld, but given %zu.", task_context.GetNodeName(), i, expected_size, input_tensor->GetSize()); REPORT_INNER_ERROR("E19999", "[%s] Input[%d]: tensor size mismatches. expected: %ld, but given %zu when %s.", - task_context.GetNodeName(), i, expected_size, input_tensor->GetSize(), __FUNCTION__); + task_context.GetNodeName(), i, expected_size, input_tensor->GetSize(), __FUNCTION__); return INTERNAL_ERROR; } } diff --git a/ge/hybrid/executor/worker/shape_inference_engine.cc b/ge/hybrid/executor/worker/shape_inference_engine.cc index f8fbbb9c..38c0fdec 100755 --- a/ge/hybrid/executor/worker/shape_inference_engine.cc +++ b/ge/hybrid/executor/worker/shape_inference_engine.cc @@ -205,8 +205,8 @@ Status ShapeInferenceEngine::UpdatePeerNodeShape(const Node &node) { auto peer_input_desc = peer_op_desc->MutableInputDesc(peer_anchor->GetIdx()); if (peer_input_desc == nullptr) { GELOGE(GRAPH_FAILED, "[Call][MutableInputDesc] for %s return nullptr.", peer_op_desc->GetName().c_str()); - REPORT_CALL_ERROR("E19999", "%s call MutableInputDesc return nullptr when ShapeInferenceEngine %s.", - peer_op_desc->GetName().c_str(), __FUNCTION__); + REPORT_CALL_ERROR("E19999", "%s call MutableInputDesc return nullptr when ShapeInferenceEngine %s.", + peer_op_desc->GetName().c_str(), __FUNCTION__); continue; } @@ -231,11 +231,11 @@ Status ShapeInferenceEngine::CanonicalizeShape(GeTensorDesc &tensor_desc, const auto &tensor_shape = tensor_desc.MutableShape(); if (tensor_shape.IsUnknownShape()) { if (!fallback_with_range) { - GELOGE(INTERNAL_ERROR, - "[Is][UnknownShape] Output shape is still unknown after shape inference. shape = [%s].", + GELOGE(INTERNAL_ERROR, + "[Is][UnknownShape] Output shape is still unknown after shape inference. shape = [%s].", tensor_shape.ToString().c_str()); REPORT_INNER_ERROR("E19999", "Output shape is still unknown after shape inference. " - "shape = [%s] when ShapeInferenceEngine %s.", tensor_shape.ToString().c_str(), __FUNCTION__); + "shape = [%s] when ShapeInferenceEngine %s.", tensor_shape.ToString().c_str(), __FUNCTION__); return INTERNAL_ERROR; } @@ -243,10 +243,10 @@ Status ShapeInferenceEngine::CanonicalizeShape(GeTensorDesc &tensor_desc, std::vector> shape_range; GE_CHK_GRAPH_STATUS_RET(tensor_desc.GetShapeRange(shape_range), "Failed to get shape range"); if (shape_range.size() != shape.size()) { - GELOGE(INTERNAL_ERROR, "[Check][Size] Number of shape ranges (%zu) mismatches that of dims (%zu).", + GELOGE(INTERNAL_ERROR, "[Check][Size] Number of shape ranges (%zu) mismatches that of dims (%zu).", shape_range.size(), shape.size()); - REPORT_INNER_ERROR("E19999", "Number of shape ranges (%zu) mismatches that of dims (%zu)" - " when ShapeInferenceEngine %s.", shape_range.size(), shape.size(), __FUNCTION__); + REPORT_INNER_ERROR("E19999", "Number of shape ranges (%zu) mismatches that of dims (%zu)" + " when ShapeInferenceEngine %s.", shape_range.size(), shape.size(), __FUNCTION__); return INTERNAL_ERROR; } @@ -270,10 +270,10 @@ Status ShapeInferenceEngine::CalcTensorSize(DataType data_type, GELOGD("To calc tensor size by shape = [%s]", GeShape(shape).ToString().c_str()); uint32_t type_size; if (!TypeUtils::GetDataTypeLength(data_type, type_size)) { - GELOGE(INTERNAL_ERROR, "[Get][DataTypeLength] failed for type:%s.", + GELOGE(INTERNAL_ERROR, "[Get][DataTypeLength] failed for type:%s.", TypeUtils::DataTypeToSerialString(data_type).c_str()); - REPORT_CALL_ERROR("E19999", "GetDataTypeLength failed for type:%s when ShapeInferenceEngine %s.", - TypeUtils::DataTypeToSerialString(data_type).c_str(), __FUNCTION__); + REPORT_CALL_ERROR("E19999", "GetDataTypeLength failed for type:%s when ShapeInferenceEngine %s.", + TypeUtils::DataTypeToSerialString(data_type).c_str(), __FUNCTION__); return INTERNAL_ERROR; } @@ -300,23 +300,24 @@ Status ShapeInferenceEngine::CalcOutputTensorSizes(const NodeItem &node_item, bo const auto &shape = tensor_desc->MutableShape(); // modify on copy auto dims = shape.GetDims(); - auto _status = CanonicalizeShape(*tensor_desc, dims, fallback_with_range); - if(_status != SUCCESS){ - REPORT_CALL_ERROR("E19999", "Invoke CanonicalizeShape failed when ShapeInferenceEngine %s, node:%s, output:%zu.", - node_item.NodeName().c_str(), __FUNCTION__, output_index); + auto status_result = CanonicalizeShape(*tensor_desc, dims, fallback_with_range); + if (status_result != SUCCESS) { + REPORT_CALL_ERROR("E19999", + "Invoke CanonicalizeShape failed when ShapeInferenceEngine %s, node:%s, output:%zu.", + node_item.NodeName().c_str(), __FUNCTION__, output_index); GELOGE(ge::FAILED, "[Canonicalize][Shape] failed for [%s], output %zu.", - node_item.NodeName().c_str(), output_index); - return _status; + node_item.NodeName().c_str(), output_index); + return status_result; } int64_t tensor_size; - _status = CalcTensorSize(tensor_desc->GetDataType(), dims, tensor_size); - if(_status != SUCCESS){ + status_result = CalcTensorSize(tensor_desc->GetDataType(), dims, tensor_size); + if (status_result != SUCCESS) { REPORT_CALL_ERROR("E19999", "Invoke CalcTensorSize failed when ShapeInferenceEngine %s, node:%s, output:%zu.", node_item.NodeName().c_str(), __FUNCTION__, output_index); GELOGE(ge::FAILED, "[Calc][TensorSize] failed for [%s], output %zu.", - node_item.NodeName().c_str(), output_index); - return _status; - } + node_item.NodeName().c_str(), output_index); + return status_result; + } GELOGD("[%s] Tensor size of output %zu = %ld", node_item.NodeName().c_str(), output_index, tensor_size); (void) TensorUtils::SetSize(*tensor_desc, tensor_size); } diff --git a/ge/hybrid/model/graph_item.cc b/ge/hybrid/model/graph_item.cc index 8204b1c8..09e0a117 100644 --- a/ge/hybrid/model/graph_item.cc +++ b/ge/hybrid/model/graph_item.cc @@ -95,8 +95,8 @@ Status GraphItem::GroupNodes() { int group = node->group; if (group != last_group) { if (seen_groups.find(group) != seen_groups.end()) { - GELOGE(INTERNAL_ERROR, - "[Order][NodeGroup]Unordered node group found. node = %s, group = %d", node->NodeName().c_str(), group); + GELOGE(INTERNAL_ERROR, + "[Find][Group]Unordered node group found. node = %s, group = %d", node->NodeName().c_str(), group); return INTERNAL_ERROR; } else { last_group = group; diff --git a/ge/hybrid/model/hybrid_model_builder.cc b/ge/hybrid/model/hybrid_model_builder.cc index e18ba2ca..ff107fe6 100755 --- a/ge/hybrid/model/hybrid_model_builder.cc +++ b/ge/hybrid/model/hybrid_model_builder.cc @@ -71,8 +71,10 @@ Status SetOutputNameAttr(ComputeGraph &graph) { } } GE_CHK_BOOL_EXEC(ge::AttrUtils::SetListStr(&graph, ATTR_MODEL_OUT_NODES_NAME, output_names), - GELOGE(FAILED, "[Invoke][SetListStr] of ATTR_MODEL_OUT_NODES_NAME failed."); - return FAILED); + GELOGE(FAILED, "[Invoke][SetListStr] failed, name:%s.", ATTR_MODEL_OUT_NODES_NAME.c_str()); + REPORT_CALL_ERROR("E19999", "SetListStr failed when %s, name:%s.", + __FUNCTION__, ATTR_MODEL_OUT_NODES_NAME.c_str()); + return FAILED); return SUCCESS; } @@ -109,10 +111,11 @@ Status CollectDependenciesForFusedGraph(NodeItem &node_item, std::set GE_CHECK_NOTNULL(src_op_desc); if (src_node->GetType() != DATA_TYPE) { GELOGE(UNSUPPORTED, - "[Check][NodeType][%s::%s] Node in fused subgraph can only depend on Data nodes, but depend on %s", + "[Check][NodeType][%s::%s] Node in fused subgraph can only depend on Data nodes," + "but depend on %s actually", node_item.NodeName().c_str(), node->GetName().c_str(), src_node->GetType().c_str()); - REPORT_INNER_ERROR("E19999", - "[%s::%s] Node in fused subgraph can only depend on Data nodes, but depend on %s when %s.", + REPORT_INNER_ERROR("E19999", "[%s::%s] Node in fused subgraph can only depend on Data nodes," + " but depend on %s actually, check invalid when %s.", node_item.NodeName().c_str(), node->GetName().c_str(), src_node->GetType().c_str(), __FUNCTION__); return UNSUPPORTED; } @@ -134,12 +137,13 @@ Status HybridModelBuilder::Build() { hybrid_model_.model_name_ = ge_root_model_->GetRootGraph()->GetName(); GELOGI("[%s] Start to build hybrid model.", GetGraphName()); GE_CHK_STATUS_RET(InitRuntimeParams(), "[Invoke][InitRuntimeParams] failed, model_name_:[%s]", GetGraphName()); - GE_CHK_STATUS_RET(RecoverGraphUnknownFlag(), "[Invoke][RecoverGraphUnknownFlag] failed, model_name_:[%s]", GetGraphName()); + GE_CHK_STATUS_RET(RecoverGraphUnknownFlag(), + "[Invoke][RecoverGraphUnknownFlag] failed, model_name_:[%s]", GetGraphName()); GE_CHK_STATUS_RET(IndexSpecialNodes(), "[Invoke][IndexSpecialNodes] failed, model_name_:[%s]", GetGraphName()); GE_CHK_STATUS_RET(IndexTaskDefs(), "[Invoke][IndexTaskDefs] failed, model_name_:[%s]", GetGraphName()); GE_CHK_STATUS_RET(InitWeights(), "[Invoke][InitWeights] failed, model_name_:[%s]", GetGraphName()); GE_CHK_STATUS_RET(LoadGraph(), "[Invoke][LoadGraph] failed, model_name_:[%s]", GetGraphName()); - GE_CHK_STATUS_RET(AssignUninitializedConstantOps(), + GE_CHK_STATUS_RET(AssignUninitializedConstantOps(), "[Invoke][AssignUninitializedConstantOps] failed, model_name_:[%s]", GetGraphName()); GE_CHK_STATUS_RET(TransAllVarData(), "[Invoke][TransAllVarData] failed, model_name_:[%s]", GetGraphName()); GE_CHK_STATUS_RET(CopyVarData(), "[Invoke][CopyVarData] failed, model_name_:[%s]", GetGraphName()); @@ -335,6 +339,8 @@ Status HybridModelBuilder::ParseDependentInputNodes(NodeItem &node_item, const s if (input_index < 0) { GELOGE(INTERNAL_ERROR, "[Get][InputIndex]failed, node:[%s] inputname: %s.", node_item.NodeName().c_str(), input_name.c_str()); + REPORT_CALL_ERROR("E19999", "GetInputIndexByName failed when HybridModelBuilder %s, node:[%s] inputname: %s.", + __FUNCTION__, node_item.NodeName().c_str(), input_name.c_str()); return INTERNAL_ERROR; } @@ -423,20 +429,24 @@ Status HybridModelBuilder::UpdateAnchorStatus(const NodePtr &node) { auto peer_anchor = anchor->GetPeerOutAnchor(); if (peer_anchor == nullptr) { if (AnchorUtils::SetStatus(anchor, ANCHOR_SUSPEND) != GRAPH_SUCCESS) { - GELOGE(INTERNAL_ERROR, "[Invoke][SetStatus] failed, node:[%s].", node->GetName().c_str()); - REPORT_CALL_ERROR("E19999", "SetStatus failed, node:[%s] when %s.", node->GetName().c_str(), __FUNCTION__); + GELOGE(INTERNAL_ERROR, "[Invoke][SetStatus] failed to set ANCHOR_SUSPEND, node:[%s].", + node->GetName().c_str()); + REPORT_CALL_ERROR("E19999", "SetStatus failed to set ANCHOR_SUSPEND, node:[%s] when HybridModelBuilder %s.", + node->GetName().c_str(), __FUNCTION__); return INTERNAL_ERROR; } } else if (peer_anchor->GetOwnerNode()->GetType() == CONSTANT) { if (AnchorUtils::SetStatus(anchor, ANCHOR_CONST) != GRAPH_SUCCESS) { - GELOGE(INTERNAL_ERROR, "[Invoke][SetStatus] failed, node:[%s].", node->GetName().c_str()); - REPORT_CALL_ERROR("E19999", "SetStatus failed, node:[%s] when %s.", node->GetName().c_str(), __FUNCTION__); + GELOGE(INTERNAL_ERROR, "[Invoke][SetStatus] failed to set ANCHOR_CONST, node:[%s].", node->GetName().c_str()); + REPORT_CALL_ERROR("E19999", "SetStatus failed to set ANCHOR_CONST, node:[%s] when HybridModelBuilder %s.", + node->GetName().c_str(), __FUNCTION__); return INTERNAL_ERROR; } } else { if (AnchorUtils::SetStatus(anchor, ANCHOR_DATA) != GRAPH_SUCCESS) { - GELOGE(INTERNAL_ERROR, "[Invoke][SetStatus] failed, node:[%s].", node->GetName().c_str()); - REPORT_CALL_ERROR("E19999", "SetStatus failed, node:[%s] when %s.", node->GetName().c_str(), __FUNCTION__); + GELOGE(INTERNAL_ERROR, "[Invoke][SetStatus] failed to set ANCHOR_DATA, node:[%s].", node->GetName().c_str()); + REPORT_CALL_ERROR("E19999", "SetStatus failed to set ANCHOR_DATA, node:[%s] when HybridModelBuilder %s.", + node->GetName().c_str(), __FUNCTION__); return INTERNAL_ERROR; } } @@ -563,6 +573,8 @@ Status HybridModelBuilder::MergeNetOutputNode(ComputeGraph &graph) { if (input_desc == nullptr) { GELOGE(INTERNAL_ERROR, "[Invoke][MutableInputDesc][%s] Failed to get input desc[%d]", net_output_desc->GetName().c_str(), index); + REPORT_CALL_ERROR("E19999", "[%s] Failed to get input desc[%d] when HybridModelBuilder %s.", + net_output_desc->GetName().c_str(), index, __FUNCTION__); return INTERNAL_ERROR; } @@ -763,7 +775,7 @@ Status HybridModelBuilder::LoadGraph() { GELOGI("Before merging subgraphs DirectNodesSize = %zu, GetAllNodesSize = %zu", root_graph->GetDirectNodesSize(), root_graph->GetAllNodesSize()); - GE_CHK_GRAPH_STATUS_RET(UnfoldSubgraphs(root_graph, merged_graph), + GE_CHK_GRAPH_STATUS_RET(UnfoldSubgraphs(root_graph, merged_graph), "[Invoke][UnfoldSubgraphs]Failed to unfold subgraphs, model_name_:%s.", GetGraphName()); root_graph = std::move(merged_graph); GELOGI("After merging subgraphs DirectNodesSize = %zu, GetAllNodesSize = %zu", @@ -786,7 +798,7 @@ Status HybridModelBuilder::LoadGraph() { op_desc->SetId(index++); } GE_DUMP(root_graph, "hybrid_merged_graph"); - GE_CHK_STATUS_RET(LoadDynamicSubgraph(*root_graph, true), + GE_CHK_STATUS_RET(LoadDynamicSubgraph(*root_graph, true), "[Invoke][LoadDynamicSubgraph]Failed to load root graph, model_name_:%s.", GetGraphName()); GELOGD("Done loading root graph successfully."); GE_CHK_STATUS_RET(hybrid_model_.root_graph_item_->GroupNodes(), @@ -825,7 +837,7 @@ Status HybridModelBuilder::LoadGraph() { } } - GE_CHK_STATUS_RET(ParseDependentByParallelGroup(), + GE_CHK_STATUS_RET(ParseDependentByParallelGroup(), "[Invoke][ParseDependentByParallelGroup]Failed to establish dependencies for hccl ops, model_name_:%s.", GetGraphName()); GELOGI("Done loading all subgraphs successfully."); @@ -941,7 +953,7 @@ Status HybridModelBuilder::InitConstantOps() { auto op_desc = var_node->GetOpDesc(); auto v_weights = ModelUtils::GetWeights(op_desc); if (v_weights.empty()) { - GELOGE(INTERNAL_ERROR, "[Check][Size][%s] Constant no not have value", var_node->GetName().c_str()); + GELOGE(INTERNAL_ERROR, "[Check][Size][%s] Constant op has no weight", var_node->GetName().c_str()); return INTERNAL_ERROR; } auto *ge_tensor = const_cast(v_weights[0].get()); @@ -955,7 +967,7 @@ Status HybridModelBuilder::InitConstantOps() { GELOGD("Init tensor with host constant %s size = %zu", var_name.c_str(), aligned_tensor.MutableData().GetSize()); if (MemManager::Instance().HostMemInstance(RT_MEMORY_HBM).Malloc(aligned_tensor.GetAlignedPtr(), aligned_tensor.GetData().size()) == nullptr) { - GELOGE(MEMALLOC_FAILED, "Malloc host memory for an existed GeTensor failed, model_name_:%s.", GetGraphName()); + GELOGE(MEMALLOC_FAILED, "[Malloc][HostMemory] for an existed GeTensor failed, model_name_:%s.", GetGraphName()); return MEMALLOC_FAILED; } var_tensor.reset(new(std::nothrow)TensorValue(aligned_tensor.MutableData().data(), @@ -1006,6 +1018,8 @@ Status HybridModelBuilder::InitVariableTensors() { int64_t tensor_size = 0; if (TensorUtils::CalcTensorMemSize(output_tensor.GetShape(), output_tensor.GetFormat(), output_tensor.GetDataType(), tensor_size) != SUCCESS) { + REPORT_CALL_ERROR("E19999", "CalcTensorMemSize failed when HybridModelBuilder %s, node name:%s", + __FUNCTION__, it.first.c_str()); GELOGE(INTERNAL_ERROR, "[Calculate][TensorMemSize] failed, node name:%s", it.first.c_str()); return INTERNAL_ERROR; } @@ -1016,7 +1030,7 @@ Status HybridModelBuilder::InitVariableTensors() { } if (MemManager::Instance().HostMemInstance(RT_MEMORY_HBM).Malloc(mem_info.host_aligned_ptr, tensor_size) == nullptr) { - GELOGE(MEMALLOC_FAILED, + GELOGE(MEMALLOC_FAILED, "[Malloc][HostMem] for an existed GeTensor failed, Host variable [%s].", it.first.c_str()); return MEMALLOC_FAILED; } @@ -1078,7 +1092,7 @@ Status HybridModelBuilder::InitWeights() { const GeTensorDesc &tensor_desc = ge_tensor->GetTensorDesc(); int64_t tensor_size = 0; GE_CHK_GRAPH_STATUS_RET(TensorUtils::GetSize(*op_desc->MutableOutputDesc(0), tensor_size), - "[Invoke][GetSize][%s] Failed to get tensor size", + "[Invoke][GetSize][%s] Failed to get output tensor size", node->GetName().c_str()); int64_t data_offset = 0; GE_CHK_GRAPH_STATUS_RET(TensorUtils::GetDataOffset(tensor_desc, data_offset), @@ -1333,16 +1347,18 @@ Status HybridModelBuilder::GetPeerNodeAcrossSubGraphs(const NodePtr &data_node, GELOGD("To get peer node of %s::%s", sub_graph->GetName().c_str(), data_node->GetName().c_str()); auto wrapped_node = data_node->GetOwnerComputeGraph()->GetParentNode(); if (wrapped_node == nullptr) { + REPORT_INNER_ERROR("E19999", "[%s] Node is in root graph when HybridModelBuilder %s.", + data_node->GetName().c_str(), __FUNCTION__); GELOGE(INTERNAL_ERROR, "[Invoke][GetParentNode][%s] Node is in root graph.", data_node->GetName().c_str()); return INTERNAL_ERROR; } auto data_op_desc = data_node->GetOpDesc(); uint32_t parent_index = 0; if (!AttrUtils::GetInt(data_op_desc, ATTR_NAME_PARENT_NODE_INDEX, parent_index)) { - GELOGE(INTERNAL_ERROR, - "[Invoke][GetInt][%s] Failed to get attr [%s]", - data_op_desc->GetName().c_str(), - ATTR_NAME_PARENT_NODE_INDEX.c_str()); + REPORT_CALL_ERROR("E19999", "[%s] Failed to get attr [%s] when HybridModelBuilder %s.", + data_op_desc->GetName().c_str(), ATTR_NAME_PARENT_NODE_INDEX.c_str(), __FUNCTION__); + GELOGE(INTERNAL_ERROR, "[Invoke][GetInt][%s] Failed to get attr [%s]", + data_op_desc->GetName().c_str(), ATTR_NAME_PARENT_NODE_INDEX.c_str()); return INTERNAL_ERROR; } @@ -1350,7 +1366,9 @@ Status HybridModelBuilder::GetPeerNodeAcrossSubGraphs(const NodePtr &data_node, GE_CHECK_NOTNULL(wrapped_node_in_anchor); auto src_out_anchor = wrapped_node_in_anchor->GetPeerOutAnchor(); if (src_out_anchor == nullptr || src_out_anchor->GetOwnerNode() == nullptr) { - GELOGE(INTERNAL_ERROR, + REPORT_INNER_ERROR("E19999", "[%s] Parent node do not have peer anchor when HybridModelBuilder %s.", + data_node->GetName().c_str(), __FUNCTION__); + GELOGE(INTERNAL_ERROR, "[Check][ParentNode][%s] Parent node do not have peer anchor.", data_node->GetName().c_str()); return INTERNAL_ERROR; } @@ -1374,10 +1392,13 @@ Status HybridModelBuilder::GetPeerNodeAcrossSubGraphs(const NodePtr &data_node, auto src_graph = NodeUtils::GetSubgraph(*src_wrapped_node, kSubgraphIndex); GE_CHECK_NOTNULL(src_graph); auto src_net_output_node = src_graph->FindFirstNodeMatchType(NETOUTPUT); - GE_CHK_BOOL_TRUE_EXEC_WITH_LOG(src_net_output_node == nullptr, - return INTERNAL_ERROR, - "[Invoke][FindFirstNodeMatchType]Failed to find NetOutput in subgraph: %s", - src_graph->GetName().c_str()); + if (src_net_output_node == nullptr) { + REPORT_INNER_ERROR("E19999", "Failed to find NetOutput in subgraph: %s when HybridModelBuilder %s", + src_graph->GetName().c_str(), __FUNCTION__); + GELOGE(INTERNAL_ERROR, "[Invoke][FindFirstNodeMatchType]Failed to find NetOutput in subgraph: %s", + src_graph->GetName().c_str()); + return INTERNAL_ERROR; + } auto net_output_desc = src_net_output_node->GetOpDesc(); GE_CHECK_NOTNULL(net_output_desc); @@ -1567,10 +1588,10 @@ Status HybridModelBuilder::GetParentNodeOutputIndex(const OpDesc &op_desc, int i auto input_desc = op_desc.MutableInputDesc(index); GE_CHECK_NOTNULL(input_desc); if (!AttrUtils::GetInt(input_desc, ATTR_NAME_PARENT_NODE_INDEX, out_index)) { - GELOGE(INTERNAL_ERROR, "[Invoke][GetInt]NetOutput input tensor %d, attr %s not found.", - index, ATTR_NAME_PARENT_NODE_INDEX.c_str()); - REPORT_CALL_ERROR("E19999", "NetOutput input tensor %d, attr %s not found when %s.", - index, ATTR_NAME_PARENT_NODE_INDEX.c_str(), __FUNCTION__); + GELOGE(INTERNAL_ERROR, "[Invoke][GetInt]NetOutput %s input tensor %d, attr %s not found.", + op_desc.GetName().c_str(), index, ATTR_NAME_PARENT_NODE_INDEX.c_str()); + REPORT_CALL_ERROR("E19999", "NetOutput %s input tensor %d, attr %s not found when %s.", + op_desc.GetName().c_str(), index, ATTR_NAME_PARENT_NODE_INDEX.c_str(), __FUNCTION__); return INTERNAL_ERROR; } return SUCCESS; @@ -2026,10 +2047,10 @@ Status HybridModelBuilder::BuildInputMapping(GraphItem &graph_item, data_op_index++; } else { if (!AttrUtils::GetInt(node->GetOpDesc(), ATTR_NAME_PARENT_NODE_INDEX, data_index)) { - GELOGE(FAILED, - "[Invoke][GetInt][%s] Failed to get attr [%s]", - node->GetName().c_str(), - ATTR_NAME_PARENT_NODE_INDEX.c_str()); + GELOGE(FAILED, "[Invoke][GetInt][%s] Failed to get attr [%s]", + node->GetName().c_str(), ATTR_NAME_PARENT_NODE_INDEX.c_str()); + REPORT_CALL_ERROR("E19999", "call GetInt failed when HybridModelBuilder %s, [%s] Failed to get attr [%s]", + __FUNCTION__, node->GetName().c_str(), ATTR_NAME_PARENT_NODE_INDEX.c_str()); return FAILED; } }