tracer.cc 8.8 KB
Newer Older
J
Jiabin Yang 已提交
1
// Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
2 3 4 5 6 7 8 9 10 11 12 13 14
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "paddle/fluid/imperative/tracer.h"
M
minqiyang 已提交
15
#include <unordered_set>
16
#include <utility>
C
chengduo 已提交
17
#include "paddle/fluid/platform/profiler.h"
M
minqiyang 已提交
18

19
namespace paddle {
M
minqiyang 已提交
20 21
namespace imperative {

J
Jiabin Yang 已提交
22 23 24 25 26 27 28 29 30 31
static std::vector<std::unique_ptr<framework::OpDesc>> CreateGradOpDescs(
    const framework::OpInfo& op_info, const framework::OpDesc& op_desc,
    const std::unordered_set<std::string>& no_grad_set,
    const std::vector<framework::BlockDesc*>& grad_sub_block,
    std::unordered_map<std::string, std::string>* grad_to_var) {
  if (op_info.grad_op_maker_) {
    return op_info.grad_op_maker_(op_desc, no_grad_set, grad_to_var,
                                  grad_sub_block);
  } else {
    return {};
X
Xin Pan 已提交
32
  }
M
minqiyang 已提交
33 34
}

J
Jiabin Yang 已提交
35 36 37 38 39 40 41 42 43 44 45 46 47
void Tracer::TraceOp(const std::string& type, const NameVarBaseMap& ins,
                     const NameVarBaseMap& outs, framework::AttributeMap attrs,
                     const platform::Place& place, bool trace_backward) {
  platform::RecordEvent event(type);
  VLOG(1) << "Trace Op: " << type;
  size_t op_id = GenerateUniqueId();
  auto op = OpBase::Create(op_id, type, ins, outs, std::move(attrs), place);
  op->Run(ins, outs);

  if (ComputeRequiredGrad(ins, outs, trace_backward)) {
    TraceBackward(op, framework::OpDesc(op->Type(), op->InputNameMap(),
                                        op->OutputNameMap(), op->Attrs()),
                  ins, outs);
48
    VLOG(6) << "Finish tracking Backward of op: " << type;
49
  }
50
  VLOG(6) << "Finish tracing fwd op: " << type;
M
minqiyang 已提交
51 52
}

J
Jiabin Yang 已提交
53
bool Tracer::ComputeRequiredGrad(const NameVarBaseMap& ins,
54
                                 const NameVarBaseMap& outs,
J
Jiabin Yang 已提交
55 56 57
                                 bool trace_backward) {
  // TODO(jiabin): Implement auto prune here
  return trace_backward;
M
minqiyang 已提交
58 59
}

J
Jiabin Yang 已提交
60 61 62 63 64 65 66
void Tracer::TraceBackward(const std::shared_ptr<OpBase>& fwd_op,
                           const framework::OpDesc& fwd_op_desc,
                           const NameVarBaseMap& ins,
                           const NameVarBaseMap& outs) {
  // grad_to_var is a map of framework::GradVarName(in_var_name/out_var_name) ->
  // in_var_name/out_var_name
  std::unordered_map<std::string, std::string> grad_to_var;
67

J
Jiabin Yang 已提交
68 69 70
  // Get grad_op_desc using fwd_op_desc
  std::vector<std::unique_ptr<framework::OpDesc>> grad_op_descs_ =
      CreateGradOpDescs(fwd_op->Info(), fwd_op_desc, {}, {}, &grad_to_var);
71

J
Jiabin Yang 已提交
72
  // Create grad_ops using grad_op_descs
73

J
Jiabin Yang 已提交
74
  size_t grad_op_num = grad_op_descs_.size();
75

J
Jiabin Yang 已提交
76 77
  VLOG(3) << "Create " << grad_op_num << " grad op desc(s) to op "
          << fwd_op->Type();
78

J
Jiabin Yang 已提交
79 80
  if (grad_op_num == 0) {
    return;
81
  }
J
Jiabin Yang 已提交
82 83 84 85 86 87 88 89 90 91
  // Build a map to record var_name -> std::shared_ptr<VarBase>*,
  // so that we can find suitable var in grad op descs
  std::unordered_map<std::string, const std::shared_ptr<VarBase>*> name_to_var;
  for (auto& pair : ins) {
    for (auto& var : pair.second) {
      auto& var_ptr = name_to_var[var->Name()];
      PADDLE_ENFORCE_EQ(var_ptr == nullptr || var_ptr->get() == var.get(), true,
                        "There are different variables with same name %s",
                        var->Name());
      var_ptr = &var;
M
minqiyang 已提交
92 93 94
    }
  }

J
Jiabin Yang 已提交
95 96 97 98 99 100 101
  for (auto& pair : outs) {
    for (auto& var : pair.second) {
      auto& var_ptr = name_to_var[var->Name()];
      PADDLE_ENFORCE_EQ(var_ptr == nullptr || var_ptr->get() == var.get(), true,
                        "There are different variables with same name %s",
                        var->Name());
      var_ptr = &var;
M
minqiyang 已提交
102 103 104
    }
  }

J
Jiabin Yang 已提交
105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152
  // Build backward ins and outs

  for (size_t i = 0; i < grad_op_num; i++) {
    // Step1: build grad op and add them to engine

    // Use trace id to decide the order of gradient sum in sorted sum mode
    size_t trace_id = fwd_op->id();
    std::shared_ptr<OpBase> grad_op =
        OpBase::Create(trace_id, (*(grad_op_descs_[i].get())), fwd_op->place());

    // this OpBase* is just used to manage op's life time
    engine_->InsertOp(grad_op.get(), grad_op);

    std::unordered_set<OpBase*> visited_preceding_ops;
    // Step2 : prepare grad_in vars and bind them with grad_op,
    // set inputs' grad_op as current grad_op
    for (const auto& grad_ins : grad_op_descs_[i]->Inputs()) {
      if (grad_ins.second.empty()) continue;
      auto& bwd_in = (*grad_op->GetMutableInsMap())[grad_ins.first];
      bwd_in.reserve(grad_ins.second.size());

      for (auto& grad_in_var_name : grad_ins.second) {
        auto iter = grad_to_var.find(grad_in_var_name);

        if (iter != grad_to_var.end()) {
          // If it is a grad var, find its coresponding forward var
          auto& fwd_var_name = iter->second;
          auto fwd_var_iter = name_to_var.find(fwd_var_name);
          PADDLE_ENFORCE_EQ(fwd_var_iter != name_to_var.end(), true,
                            "Cannot find forward variable named %s",
                            fwd_var_name);
          PADDLE_ENFORCE_NOT_NULL(
              (*(fwd_var_iter->second))->GradVarBase(),
              "Grad of %s should "
              "not be NULL when we Track_Backward Input of %s",
              (*(fwd_var_iter->second))->Name(), grad_op->Type());
          (*(fwd_var_iter->second))->GradVarBase()->AddGradOps(grad_op);
          VLOG(3) << "Add Grad Op " << grad_op->Type() << " for :"
                  << (*(fwd_var_iter->second))->GradVarBase()->Name();
          bwd_in.emplace_back((*(fwd_var_iter->second))->GradVarBase());
        } else {
          // If it is a forward var, just add it
          auto fwd_var_iter = name_to_var.find(grad_in_var_name);
          PADDLE_ENFORCE_EQ(fwd_var_iter != name_to_var.end(), true,
                            "Cannot find forward variable named %s",
                            grad_in_var_name);
          bwd_in.emplace_back(*(fwd_var_iter->second));
        }
153

J
Jiabin Yang 已提交
154 155 156 157 158
        VLOG(3) << "Set backward input " << grad_ins.first << " of "
                << grad_op->Type() << " to be "
                << (bwd_in.back() ? bwd_in.back()->Name() : "nullptr");
      }
    }
159

J
Jiabin Yang 已提交
160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194
    // Step3: prepare grad_out vars and using their grad_ops to set current
    // grad_op's preceding op
    for (auto& grad_outs : grad_op_descs_[i]->Outputs()) {
      if (grad_outs.second.empty()) continue;
      auto& bwd_out = (*grad_op->GetMutableOutsMap())[grad_outs.first];
      bwd_out.reserve(grad_outs.second.size());

      for (auto& grad_out_var_name : grad_outs.second) {
        auto iter = grad_to_var.find(grad_out_var_name);
        PADDLE_ENFORCE_EQ(iter != grad_to_var.end(), true,
                          "Cannot find output of input grad %s in op %s",
                          grad_out_var_name, fwd_op->Type());
        auto fwd_var_iter = name_to_var.find(iter->second);
        PADDLE_ENFORCE_EQ(fwd_var_iter != name_to_var.end(), true,
                          "Cannot find forward variable named %s",
                          iter->second);
        PADDLE_ENFORCE_NOT_NULL(
            (*(fwd_var_iter->second))->GradVarBase(),
            "Grad of %s should "
            "not be NULL when we Track_Backward Output of %s",
            (*(fwd_var_iter->second))->Name(), grad_op->Type());
        bwd_out.emplace_back((*(fwd_var_iter->second))->GradVarBase());
        VLOG(3) << "Set backward output " << grad_outs.first << " of "
                << grad_op->Type() << " to be "
                << (bwd_out.back() ? bwd_out.back()->Name() : "nullptr");

        auto preceding_ops =
            (*(fwd_var_iter->second))->GradVarBase()->GradOps();

        if (VLOG_IS_ON(3) && !preceding_ops.empty()) {
          VLOG(3) << "Add preceding Op of :"
                  << (*(fwd_var_iter->second))->GradVarBase()->Name()
                  << " It's preceding Op are: ";
          for (const auto& op : preceding_ops) {
            VLOG(3) << op->Type();
X
Xin Pan 已提交
195 196
          }
        }
197

J
Jiabin Yang 已提交
198 199 200 201 202 203 204 205 206 207 208 209
        if (!preceding_ops.empty()) {
          for (const auto& op : preceding_ops) {
            PADDLE_ENFORCE_NOT_NULL(op, "No nullptr should be preceding_op");
            if (visited_preceding_ops.count(op) == 0) {
              visited_preceding_ops.insert(op);
              grad_op->InsertGradPendingOps(op);
            }
          }
        } else {
          VLOG(5) << "Hit leaf VarBase";
          VLOG(5) << "Hit leaf VarBase"
                  << (*(fwd_var_iter->second))->GradVarBase()->Name();
M
minqiyang 已提交
210 211 212
        }
      }
    }
J
Jiabin Yang 已提交
213 214
    // To ensure numeric stability as static graph
    grad_op->SortGradPendingOps();
M
minqiyang 已提交
215 216
  }
}
J
Jiabin Yang 已提交
217

M
minqiyang 已提交
218
}  // namespace imperative
219
}  // namespace paddle