softmax_with_cross_entropy_op.cc 7.8 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.

   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/operators/softmax_with_cross_entropy_op.h"
Y
Yu Yang 已提交
16
#include <paddle/function/TensorType.h>
17 18 19 20 21 22 23

namespace paddle {
namespace operators {

class SoftmaxWithCrossEntropyOpMaker
    : public framework::OpProtoAndCheckerMaker {
 public:
24 25
  SoftmaxWithCrossEntropyOpMaker(framework::OpProto* proto,
                                 framework::OpAttrChecker* op_checker)
26
      : OpProtoAndCheckerMaker(proto, op_checker) {
C
caoying03 已提交
27
    AddInput("Logits",
28
             "(Tensor, default: Tensor<float>), The unscaled log probabilities "
C
caoying03 已提交
29
             "which is a 2-D tensor with shape [N x K]. N is the batch_size, "
30 31 32 33 34 35 36 37
             "and K is the class number.");
    AddInput("Label",
             "(Tensor, default: Tensor<int>), The ground truth which is a 2-D "
             "tensor. "
             "If softLable is set to 0, Label is a Tensor<int> with shape [N x "
             "1]. "
             "If softLable is set to 1, Label is a Tensor<float/double> "
             "with shape [N x K].");
C
caoying03 已提交
38 39
    AddOutput(
        "Softmax",
40
        "(Tensor, default: Tensor<float>), A 2-D tensor with shape [N x K]. "
C
caoying03 已提交
41 42
        "The outputs value of softmax activation by given the input batch, "
        "which will be used in backward calculation.")
C
caoying03 已提交
43
        .AsIntermediate();
C
caoying03 已提交
44
    AddOutput("Loss",
45
              "(Tensor, default: Tensor<float>), A 2-D tensor. The cross "
C
caoying03 已提交
46
              "entropy loss with shape [N x 1].");
C
caoying03 已提交
47 48 49 50 51
    AddAttr<bool>(
        "softLabel",
        "(bool, default: false), A flag to indicate whether to interpretate "
        "the given labels as soft labels.")
        .SetDefault(false);
52 53 54 55 56 57 58 59 60 61 62 63 64 65
    AddComment(R"DOC(
Cross entropy loss with softmax are used as the output layer extensively. This
operator computes the softmax normalized values for each row of the input
tensor, after which cross-entropy loss is then computed. This provides a more
numerically stable gradient.

Because this operators performs a softmax on logits internally, it expects
unscaled logits. Please do not call this op with the output of softmax operator,
which will produce incorrect results.

This operators expects mutually exclusive hard labels, each sample in a batch
is in exactly one class with probabilities 1. Each sample in the batch with one
and only one label.

C
caoying03 已提交
66
Equation:
67

C
caoying03 已提交
68
1) hard label (one-hot label)
69

C
caoying03 已提交
70 71 72 73 74 75 76
Loss_j = -\text{Logit}_{Label_j} + \log\left(\sum_{i=0}^{K}\exp(\text{Logit}_i)\right), j = 1, ..., K

2) soft label (a distribution over all classes)

Loss_j = -\sum_{i=0}^{K}\text{Label}_i\left(\text{Logit}_i-\log\left(\sum_{i=0}^{K}\exp(\text{Logit}_i)\right)\right), j = 1,...,K

)DOC");
77 78 79 80 81 82 83 84
  }
};

class SoftmaxWithCrossEntropyOp : public framework::OperatorWithKernel {
 public:
  using framework::OperatorWithKernel::OperatorWithKernel;

 protected:
Q
qiaolongfei 已提交
85 86 87 88 89 90 91 92 93 94 95
  void InferShape(framework::InferShapeContextBase* ctx) const override {
    PADDLE_ENFORCE(ctx->HasInput("Logits"),
                   "Input(Logits) should be not null.");
    PADDLE_ENFORCE(ctx->HasInput("Label"), "Input(Label) should be not null.");

    PADDLE_ENFORCE(ctx->HasOutput("Softmax"),
                   "Output(Softmax) should be not null.");
    PADDLE_ENFORCE(ctx->HasOutput("Loss"), "Output(Loss) should be not null.");

    auto logits_dims = ctx->GetInputDim("Logits");
    auto labels_dims = ctx->GetInputDim("Label");
C
caoying03 已提交
96
    PADDLE_ENFORCE_EQ(
Q
qiaolongfei 已提交
97
        logits_dims.size(), 2UL,
98
        "The input of softmax_with_cross_entropy should be a 2-D tensor.");
Q
qiaolongfei 已提交
99
    PADDLE_ENFORCE_EQ(labels_dims.size(), 2UL,
C
caoying03 已提交
100
                      "The labels should be a 2-D tensor.");
101

Q
qiaolongfei 已提交
102 103
    if (ctx->Attrs().Get<bool>("softLabel")) {
      PADDLE_ENFORCE_EQ(logits_dims[1], labels_dims[1],
104 105 106
                        "If Attr(softLabel) == true, the 2nd dimension of "
                        "Input(X) and Input(Label) should be equal.");
    } else {
Q
qiaolongfei 已提交
107
      PADDLE_ENFORCE_EQ(labels_dims[1], 1UL,
108 109 110 111
                        "If Attr(softLabel) == false, the 2nd dimension of "
                        "Input(Label) should be 1.");
    }

Q
qiaolongfei 已提交
112 113
    ctx->SetOutputDim("Softmax", logits_dims);
    ctx->SetOutputDim("Loss", {logits_dims[0], 1});
114

Q
qiaolongfei 已提交
115 116
    ctx->ShareLoD("Logits", /*->*/ "Softmax");
    ctx->ShareLoD("Logits", /*->*/ "Loss");
C
caoying03 已提交
117
  }
Y
Yu Yang 已提交
118 119 120 121 122

  framework::DataType IndicateDataType(
      const framework::ExecutionContext& ctx) const override {
    return framework::ToDataType(ctx.Input<Tensor>("Logits")->type());
  }
C
caoying03 已提交
123 124 125 126 127 128 129
};

class SoftmaxWithCrossEntropyOpGrad : public framework::OperatorWithKernel {
 public:
  using framework::OperatorWithKernel::OperatorWithKernel;

 protected:
Q
qiaolongfei 已提交
130 131 132 133 134 135 136 137 138 139 140 141
  void InferShape(framework::InferShapeContextBase* ctx) const override {
    PADDLE_ENFORCE(ctx->HasInput(framework::GradVarName("Loss")),
                   "Input(Loss@Grad) should not be null.");
    PADDLE_ENFORCE(ctx->HasInput("Softmax"),
                   "Input(Softmax) should be not null.");
    PADDLE_ENFORCE(ctx->HasInput("Label"), "Input(Label) should be not null.");
    PADDLE_ENFORCE(ctx->HasOutput(framework::GradVarName("Logits")),
                   "Output(Logits@Grad) should be not null.");

    auto softmax_dims = ctx->GetInputDim("Softmax");
    auto labels_dims = ctx->GetInputDim("Label");
    PADDLE_ENFORCE_EQ(labels_dims.size(), 2UL,
C
caoying03 已提交
142
                      "The labels should be a 2-D tensor.");
143

Q
qiaolongfei 已提交
144 145
    if (ctx->Attrs().Get<bool>("softLabel")) {
      PADDLE_ENFORCE_EQ(softmax_dims[1], labels_dims[1],
146 147 148
                        "When Attr(softLabel) == true, the 2nd dimension of "
                        "Input(X) and Input(Label) should be equal.");
    } else {
Q
qiaolongfei 已提交
149
      PADDLE_ENFORCE_EQ(labels_dims[1], 1UL,
150 151 152
                        "When Attr(softLabel) == false, the 2nd dimension of "
                        "Input(Label) should be 1.");
    }
C
caoying03 已提交
153

Q
qiaolongfei 已提交
154 155
    ctx->SetOutputDim(framework::GradVarName("Logits"),
                      ctx->GetInputDim("Softmax"));
156
  }
Y
Yu Yang 已提交
157 158 159

  framework::DataType IndicateDataType(
      const framework::ExecutionContext& ctx) const override {
Y
Fix CI  
Yu Yang 已提交
160 161
    return framework::ToDataType(
        ctx.Input<Tensor>(framework::GradVarName("Loss"))->type());
Y
Yu Yang 已提交
162
  }
163 164
};

165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183
class SoftmaxGradMaker : public framework::SingleGradOpDescMaker {
 public:
  using framework::SingleGradOpDescMaker::SingleGradOpDescMaker;

 protected:
  framework::OpDescBind Apply() const override {
    framework::OpDescBind grad_op;
    grad_op.SetType("softmax_with_cross_entropy_grad");
    grad_op.SetInput("Label", Input("Label"));
    grad_op.SetInput("Softmax", Output("Softmax"));
    grad_op.SetInput("Loss", Output("Loss"));
    grad_op.SetInput(framework::GradVarName("Softmax"), OutputGrad("Softmax"));
    grad_op.SetInput(framework::GradVarName("Loss"), OutputGrad("Loss"));
    grad_op.SetOutput(framework::GradVarName("Logits"), InputGrad("Logits"));
    grad_op.SetAttrMap(Attrs());
    return grad_op;
  }
};

184 185 186 187 188
}  // namespace operators
}  // namespace paddle

namespace ops = paddle::operators;

189 190 191 192 193
REGISTER_OPERATOR(softmax_with_cross_entropy, ops::SoftmaxWithCrossEntropyOp,
                  ops::SoftmaxWithCrossEntropyOpMaker,
                  ops::SoftmaxWithCrossEntropyOpMaker);
REGISTER_OPERATOR(softmax_with_cross_entropy_grad,
                  ops::SoftmaxWithCrossEntropyOpGrad);
194 195 196 197
REGISTER_OP_CPU_KERNEL(softmax_with_cross_entropy,
                       ops::SoftmaxWithCrossEntropyKernel<float>);
REGISTER_OP_CPU_KERNEL(softmax_with_cross_entropy_grad,
                       ops::SoftmaxWithCrossEntropyGradKernel<float>);