From 7f6d8acecb0c1d61dad645c581cd8cef9d554841 Mon Sep 17 00:00:00 2001 From: peizhilin Date: Tue, 25 Dec 2018 14:12:01 +0800 Subject: [PATCH] cherry-pick the #12759 test=develop --- paddle/fluid/framework/op_proto_maker.cc | 4 ++ paddle/fluid/framework/op_proto_maker.h | 1 + paddle/fluid/framework/operator.cc | 71 ++++++++++++++----- paddle/fluid/operators/top_k_op.cc | 2 + paddle/fluid/pybind/const_value.cc | 3 + python/paddle/fluid/framework.py | 5 ++ .../tests/unittests/test_operator_desc.py | 2 +- 7 files changed, 70 insertions(+), 18 deletions(-) diff --git a/paddle/fluid/framework/op_proto_maker.cc b/paddle/fluid/framework/op_proto_maker.cc index ca31303f7..2311614c3 100644 --- a/paddle/fluid/framework/op_proto_maker.cc +++ b/paddle/fluid/framework/op_proto_maker.cc @@ -82,6 +82,10 @@ void OpProtoAndCheckerMaker::operator()(proto::OpProto* proto, AddAttr(OpNamescopeAttrName(), "Operator name with namesope.") .SetDefault(""); + AddAttr>(OpCreationCallstackAttrName(), + "Callstack for Op Creatation.") + .SetDefault({}); + Validate(); } diff --git a/paddle/fluid/framework/op_proto_maker.h b/paddle/fluid/framework/op_proto_maker.h index 4c59c73d8..0a0f8f465 100644 --- a/paddle/fluid/framework/op_proto_maker.h +++ b/paddle/fluid/framework/op_proto_maker.h @@ -47,6 +47,7 @@ class OpProtoAndCheckerMaker { static const char *OpRoleAttrName() { return "op_role"; } static const char *OpRoleVarAttrName() { return "op_role_var"; } static const char *OpNamescopeAttrName() { return "op_namescope"; } + static const char *OpCreationCallstackAttrName() { return "op_callstack"; } void operator()(proto::OpProto *proto, OpAttrChecker *attr_checker); diff --git a/paddle/fluid/framework/operator.cc b/paddle/fluid/framework/operator.cc index fec311e3e..4527e6619 100644 --- a/paddle/fluid/framework/operator.cc +++ b/paddle/fluid/framework/operator.cc @@ -16,10 +16,15 @@ limitations under the License. */ #include #include - +#include +#include +#include +#include "gflags/gflags.h" +#include "glog/logging.h" #include "paddle/fluid/framework/data_transform.h" #include "paddle/fluid/framework/executor.h" #include "paddle/fluid/framework/lod_tensor.h" +#include "paddle/fluid/framework/op_proto_maker.h" #include "paddle/fluid/framework/operator.h" #include "paddle/fluid/framework/shape_inference.h" #include "paddle/fluid/framework/transfer_scope_cache.h" @@ -157,27 +162,59 @@ RuntimeContext::RuntimeContext(const VariableNameMap& innames, } void OperatorBase::Run(const Scope& scope, const platform::Place& place) { - VLOG(4) << place << " " << DebugStringEx(&scope); - if (platform::is_gpu_place(place)) { + try { + if (VLOG_IS_ON(4)) { + VLOG(4) << place << " " << DebugStringEx(&scope); + } + if (platform::is_gpu_place(place)) { #ifndef PADDLE_WITH_CUDA - PADDLE_THROW("Cannot run operator on place %s", place); + PADDLE_THROW("Cannot run operator on place %s", place); #else - auto dev_id = boost::get(place).device; - platform::SetDeviceId(dev_id); + auto dev_id = boost::get(place).device; + platform::SetDeviceId(dev_id); #endif - } + } - // The profile has a process-wide mutex, results in serious performance issue - // in concurrency scenerio. Here use an `if` to fix this issue. - // Please not remove the `if`, ask @Superjomn if there are any concern. - if (platform::IsProfileEnabled()) { - platform::DeviceContextPool& pool = platform::DeviceContextPool::Instance(); - platform::RecordEvent record_event(Type(), pool.Get(place)); - RunImpl(scope, place); - } else { - RunImpl(scope, place); + // The profile has a process-wide mutex, results in serious performance + // issue + // in concurrency scenerio. Here use an `if` to fix this issue. + // Please not remove the `if`, ask @Superjomn if there are any concern. + if (platform::IsProfileEnabled()) { + platform::DeviceContextPool& pool = + platform::DeviceContextPool::Instance(); + platform::RecordEvent record_event(Type(), pool.Get(place)); + RunImpl(scope, place); + } else { + RunImpl(scope, place); + } + + if (VLOG_IS_ON(3)) { + VLOG(3) << place << " " << DebugStringEx(&scope); + } + } catch (platform::EnforceNotMet exception) { + if (Attrs().count("sub_block") != 0) { + throw exception; + } + + auto& callstack = Attr>( + OpProtoAndCheckerMaker::OpCreationCallstackAttrName()); + + if (callstack.empty()) { + throw exception; + } + std::ostringstream sout; + sout << "Invoke operator " << Type() << " error.\n"; + sout << "Python Callstacks: \n"; + for (auto& line : callstack) { + sout << line; + } + sout << "C++ Callstacks: \n"; + sout << exception.err_str_; + exception.err_str_ = sout.str(); + throw exception; + } catch (...) { + std::rethrow_exception(std::current_exception()); } - VLOG(3) << place << " " << DebugStringEx(&scope); } bool OperatorBase::HasInputs(const std::string& name) const { diff --git a/paddle/fluid/operators/top_k_op.cc b/paddle/fluid/operators/top_k_op.cc index c17d1afc3..e634f7c9a 100644 --- a/paddle/fluid/operators/top_k_op.cc +++ b/paddle/fluid/operators/top_k_op.cc @@ -30,6 +30,8 @@ class TopkOp : public framework::OperatorWithKernel { "Output(Indices) of TopkOp should not be null."); auto input_dims = ctx->GetInputDim("X"); + PADDLE_ENFORCE_EQ(input_dims.size(), 2, + "Rank of TopK op's input must be 2."); const int k = static_cast(ctx->Attrs().Get("k")); PADDLE_ENFORCE_GE(k, 1, "k must >= 1"); diff --git a/paddle/fluid/pybind/const_value.cc b/paddle/fluid/pybind/const_value.cc index 06d8b65fb..f8ded9f94 100644 --- a/paddle/fluid/pybind/const_value.cc +++ b/paddle/fluid/pybind/const_value.cc @@ -49,6 +49,9 @@ void BindConstValue(pybind11::module* m) { op_proto_and_checker_maker.def( "kOpNameScopeAttrName", framework::OpProtoAndCheckerMaker::OpNamescopeAttrName); + op_proto_and_checker_maker.def( + "kOpCreationCallstackAttrName", + framework::OpProtoAndCheckerMaker::OpCreationCallstackAttrName); } } // namespace pybind diff --git a/python/paddle/fluid/framework.py b/python/paddle/fluid/framework.py index de30ed2fc..3427fb0c4 100644 --- a/python/paddle/fluid/framework.py +++ b/python/paddle/fluid/framework.py @@ -20,6 +20,7 @@ import os import re import six import sys +import traceback import numpy as np @@ -604,6 +605,10 @@ class Operator(object): if role_var_name in op_attrs and len(op_attrs[role_var_name]) == 0: del op_attrs[role_var_name] + callstack_var_name = op_maker.kOpCreationCallstackAttrName() + op_attrs[callstack_var_name] = list( + reversed(traceback.format_stack()))[1:] + if len(self.desc.type()) != 0: return if type is None: diff --git a/python/paddle/fluid/tests/unittests/test_operator_desc.py b/python/paddle/fluid/tests/unittests/test_operator_desc.py index 4153394c1..37b9a9188 100644 --- a/python/paddle/fluid/tests/unittests/test_operator_desc.py +++ b/python/paddle/fluid/tests/unittests/test_operator_desc.py @@ -69,7 +69,7 @@ class TestOperator(unittest.TestCase): set(mul_op.attr_names), set([ "x_num_col_dims", "y_num_col_dims", "op_role", "op_role_var", - "op_namescope" + "op_namescope", "op_callstack" ])) self.assertEqual(mul_op.has_attr("x_num_col_dims"), True) self.assertEqual(mul_op.attr_type("x_num_col_dims"), core.AttrType.INT) -- GitLab