batch_norm_op.cc 25.7 KB
Newer Older
1
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
Q
Qiao Longfei 已提交
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. */

Y
Yi Wang 已提交
15
#include "paddle/fluid/operators/batch_norm_op.h"
Q
qingqing01 已提交
16
#include <memory>
S
Siddharth Goyal 已提交
17
#include <string>
Q
qingqing01 已提交
18
#include <unordered_map>
Y
Yi Wang 已提交
19
#include "paddle/fluid/framework/data_layout.h"
20 21 22
#ifdef PADDLE_WITH_MKLDNN
#include "paddle/fluid/platform/mkldnn_helper.h"
#endif
Q
Qiao Longfei 已提交
23 24 25 26

namespace paddle {
namespace operators {

Q
qingqing01 已提交
27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48
void BatchNormOp::InferShape(framework::InferShapeContext *ctx) const {
  PADDLE_ENFORCE(ctx->HasInput("X"), "Input(X) of ConvOp should not be null.");
  PADDLE_ENFORCE(ctx->HasInput("Scale"),
                 "Input(Scale) of ConvOp should not be null.");
  PADDLE_ENFORCE(ctx->HasInput("Bias"),
                 "Input(Bias) of ConvOp should not be null.");
  PADDLE_ENFORCE(ctx->HasInput("Mean"),
                 "Input(Mean) of ConvOp should not be null.");
  PADDLE_ENFORCE(ctx->HasInput("Variance"),
                 "Input(Variance) of ConvOp should not be null.");
  PADDLE_ENFORCE(ctx->HasOutput("Y"),
                 "Output(Y) of ConvOp should not be null.");
  bool is_test = ctx->Attrs().Get<bool>("is_test");
  if (!is_test) {
    PADDLE_ENFORCE(ctx->HasOutput("MeanOut"),
                   "Output(MeanOut) of ConvOp should not be null.");
    PADDLE_ENFORCE(ctx->HasOutput("VarianceOut"),
                   "Output(VarianceOut) of ConvOp should not be null.");
    PADDLE_ENFORCE(ctx->HasOutput("SavedMean"),
                   "Output(SavedMean) of ConvOp should not be null.");
    PADDLE_ENFORCE(ctx->HasOutput("SavedVariance"),
                   "Output(SavedVariance) of ConvOp should not be null.");
Q
Qiao Longfei 已提交
49
  }
K
Kexin Zhao 已提交
50

Q
qingqing01 已提交
51 52 53 54 55 56 57 58 59 60
  // make sure Mean/MeanOut and Variance/VarianceOut share memory in Python
  PADDLE_ENFORCE_EQ(ctx->Inputs("Mean")[0], ctx->Outputs("MeanOut")[0],
                    "Mean and MeanOut should share the same memory");
  PADDLE_ENFORCE_EQ(ctx->Inputs("Variance")[0], ctx->Outputs("VarianceOut")[0],
                    "Variance and VarianceOut should share the same memory");

  const auto x_dims = ctx->GetInputDim("X");
  const DataLayout data_layout = framework::StringToDataLayout(
      ctx->Attrs().Get<std::string>("data_layout"));

61 62 63 64 65 66 67 68 69 70 71 72
  PADDLE_ENFORCE_GE(
      x_dims.size(), 2,
      "ShapeError: the dimension of input X must greater than or equal to 2."
      "But received: the shape of input X = [%s], the dimension of input X ="
      "[%d]",
      x_dims, x_dims.size());
  PADDLE_ENFORCE_LE(
      x_dims.size(), 5,
      "ShapeError: the dimension of input X must smaller than or equal to 5."
      "But received: the shape of input X = [%s], the dimension of input X ="
      "[%d]",
      x_dims, x_dims.size());
Q
qingqing01 已提交
73 74 75 76 77

  const int64_t C =
      (data_layout == DataLayout::kNCHW ? x_dims[1]
                                        : x_dims[x_dims.size() - 1]);

78 79
  auto scale_dim = ctx->GetInputDim("Scale");
  auto bias_dim = ctx->GetInputDim("Bias");
Q
qingqing01 已提交
80

81 82 83 84 85 86 87 88 89 90
  PADDLE_ENFORCE_EQ(scale_dim.size(), 1UL,
                    "ShapeError: the dimension of scale must equal to 1."
                    "But received: the shape of scale is [%s], the dimension "
                    "of scale is [%d]",
                    scale_dim, scale_dim.size());
  PADDLE_ENFORCE_EQ(
      bias_dim.size(), 1UL,
      "ShapeError: the dimension of bias must equal to 1."
      "But received: the shape of bias is [%s],the dimension of bias is [%d]",
      bias_dim, bias_dim.size());
C
ceci3 已提交
91

92 93 94 95 96 97 98
  bool check = true;
  if ((!ctx->IsRuntime()) && (framework::product(scale_dim) <= 0 ||
                              framework::product(bias_dim) <= 0)) {
    check = false;
  }

  if (check) {
99 100 101 102 103 104 105 106
    PADDLE_ENFORCE_EQ(scale_dim[0], C,
                      "ShapeError: the shape of scale must equal to [%d]"
                      "But received: the shape of scale is [%d]",
                      C, scale_dim[0]);
    PADDLE_ENFORCE_EQ(bias_dim[0], C,
                      "ShapeError: the shape of bias must equal to [%d]"
                      "But received: the shape of bias is [%d]",
                      C, bias_dim[0]);
107
  }
Q
qingqing01 已提交
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
  ctx->SetOutputDim("Y", x_dims);
  ctx->SetOutputDim("MeanOut", {C});
  ctx->SetOutputDim("VarianceOut", {C});
  ctx->SetOutputDim("SavedMean", {C});
  ctx->SetOutputDim("SavedVariance", {C});
  ctx->ShareLoD("X", "Y");
}

framework::OpKernelType BatchNormOp::GetExpectedKernelType(
    const framework::ExecutionContext &ctx) const {
  auto input_data_type = ctx.Input<Tensor>("X")->type();
  // By default, the type of the scale, bias, mean,
  // and var tensors should both be float. (For float or float16 input tensor)
  // or double (For double input tensor).
  auto bn_param_type = framework::proto::VarType::FP32;
  if (input_data_type == framework::proto::VarType::FP64) {
    bn_param_type = framework::proto::VarType::FP64;
  }
  PADDLE_ENFORCE_EQ(bn_param_type, ctx.Input<Tensor>("Scale")->type(),
                    "Scale input should be of float type");
  PADDLE_ENFORCE_EQ(bn_param_type, ctx.Input<Tensor>("Bias")->type(),
                    "Bias input should be of float type");
  PADDLE_ENFORCE_EQ(bn_param_type, ctx.Input<Tensor>("Mean")->type(),
                    "Mean input should be of float type");
  PADDLE_ENFORCE_EQ(bn_param_type, ctx.Input<Tensor>("Variance")->type(),
                    "Variance input should be of float type");

  // TODO(pzelazko-intel): enable MKLDNN layout when it's ready
  framework::LibraryType library = framework::LibraryType::kPlain;
  framework::DataLayout layout = framework::DataLayout::kAnyLayout;
138
#ifdef PADDLE_WITH_MKLDNN
Q
qingqing01 已提交
139 140 141 142
  if (library == framework::LibraryType::kPlain &&
      platform::CanMKLDNNBeUsed(ctx)) {
    library = framework::LibraryType::kMKLDNN;
    layout = framework::DataLayout::kMKLDNN;
K
Kexin Zhao 已提交
143
  }
Q
qingqing01 已提交
144
#endif
Q
Qiao Longfei 已提交
145

Q
qingqing01 已提交
146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205
  return framework::OpKernelType(input_data_type, ctx.GetPlace(), layout,
                                 library);
}

void BatchNormOpMaker::Make() {
  AddAttr<bool>("is_test",
                "(bool, default false) Set to true for inference only, false "
                "for training. Some layers may run faster when this is true.")
      .SetDefault(false);
  AddAttr<float>("momentum", "").SetDefault(0.9);
  AddAttr<float>("epsilon", "")
      .SetDefault(1e-5)
      .AddCustomChecker([](const float &epsilon) {
        PADDLE_ENFORCE(epsilon >= 0.0f && epsilon <= 0.001f,
                       "'epsilon' should be between 0.0 and 0.001.");
      });
  AddAttr<std::string>("data_layout", "").SetDefault("NCHW");
  AddInput("X", "The input tensor");
  AddInput("Scale",
           "Scale is a 1-dimensional tensor of size C "
           "that is applied to the output");
  AddInput("Bias",
           "Bias is a 1-dimensional tensor of size C "
           "that is applied to the output");
  AddInput("Mean",
           "The global mean (for training) or "
           "estimated mean (for testing)");
  AddInput("Variance",
           "The global variance (for training) "
           "or estimated Variance (for testing)");
  AddOutput("Y", "result after normalization");
  AddOutput("MeanOut",
            "Share memory with Mean. "
            "Store the global mean when training");
  AddOutput("VarianceOut",
            "Share memory with Variance. "
            "Store the global Variance when training");
  AddOutput("SavedMean",
            "Mean of the current mini batch, "
            "will apply to output when training")
      .AsIntermediate();
  AddOutput("SavedVariance",
            "Variance of the current mini batch, "
            "will apply to output when training")
      .AsIntermediate();
  AddAttr<bool>("use_mkldnn",
                "(bool, default false) Only used in mkldnn kernel")
      .SetDefault(false);
  AddAttr<bool>("fuse_with_relu",
                "(bool, default false) Only used in mkldnn kernel")
      .SetDefault(false);
  AddAttr<bool>("use_global_stats",
                "(bool, default false) Whether to use global mean and "
                "variance. In inference or test mode, set use_global_stats "
                "to true or is_test true. the behavior is equivalent. "
                "In train mode, when setting use_global_stats True, the "
                "global mean and variance are also used during train time, "
                "the BN acts as scaling and shiffting.")
      .SetDefault(false);
  AddComment(R"DOC(
206
Batch Normalization.
Q
Qiao Longfei 已提交
207

208 209 210 211 212 213
Batch Norm has been implemented as discussed in the paper:
https://arxiv.org/pdf/1502.03167.pdf
Can be used as a normalizer function for conv2d and fully_connected operations.
The required data format for this layer is one of the following:
1. NHWC `[batch, in_height, in_width, in_channels]`
2. NCHW `[batch, in_channels, in_height, in_width]`
Q
Qiao Longfei 已提交
214 215

)DOC");
Q
qingqing01 已提交
216
}
C
chengduo 已提交
217

Q
Qiao Longfei 已提交
218
template <typename T>
Q
QI JUN 已提交
219 220
class BatchNormKernel<platform::CPUDeviceContext, T>
    : public framework::OpKernel<T> {
Q
Qiao Longfei 已提交
221 222 223 224 225
 public:
  void Compute(const framework::ExecutionContext &ctx) const override {
    const float epsilon = ctx.Attr<float>("epsilon");
    const float momentum = ctx.Attr<float>("momentum");
    const bool is_test = ctx.Attr<bool>("is_test");
226 227 228 229
    const bool use_global_stats = ctx.Attr<bool>("use_global_stats");

    bool global_stats = is_test || use_global_stats;

Q
QI JUN 已提交
230 231 232
    const std::string data_layout_str = ctx.Attr<std::string>("data_layout");
    const DataLayout data_layout =
        framework::StringToDataLayout(data_layout_str);
Q
Qiao Longfei 已提交
233 234 235

    const auto *x = ctx.Input<Tensor>("X");
    const auto &x_dims = x->dims();
236 237
    PADDLE_ENFORCE(x_dims.size() >= 2 && x_dims.size() <= 5,
                   "The Input dim size should be between 2 and 5");
Q
Qiao Longfei 已提交
238 239
    const int N = x_dims[0];
    const int C =
Q
QI JUN 已提交
240 241
        (data_layout == DataLayout::kNCHW ? x_dims[1]
                                          : x_dims[x_dims.size() - 1]);
Q
Qiao Longfei 已提交
242 243 244 245 246 247 248 249 250 251 252 253 254 255 256
    const int sample_size = x->numel() / N / C;

    auto *y = ctx.Output<Tensor>("Y");
    auto *mean_out = ctx.Output<Tensor>("MeanOut");
    auto *variance_out = ctx.Output<Tensor>("VarianceOut");
    auto *saved_mean = ctx.Output<Tensor>("SavedMean");
    auto *saved_variance = ctx.Output<Tensor>("SavedVariance");

    // alloc memory
    y->mutable_data<T>(ctx.GetPlace());
    mean_out->mutable_data<T>(ctx.GetPlace());
    variance_out->mutable_data<T>(ctx.GetPlace());
    saved_mean->mutable_data<T>(ctx.GetPlace());
    saved_variance->mutable_data<T>(ctx.GetPlace());

257
    if (!global_stats) {
Q
Qiao Longfei 已提交
258 259 260 261 262 263 264 265
      // saved_xx is use just in this batch of data
      EigenVectorArrayMap<T> saved_mean_e(
          saved_mean->mutable_data<T>(ctx.GetPlace()), C);
      EigenVectorArrayMap<T> saved_variance_e(
          saved_variance->mutable_data<T>(ctx.GetPlace()), C);
      saved_mean_e.setZero();
      saved_variance_e.setZero();

266 267 268 269 270 271
      EigenVectorArrayMap<T> running_mean_arr(
          mean_out->mutable_data<T>(ctx.GetPlace()), C);
      EigenVectorArrayMap<T> running_var_arr(
          variance_out->mutable_data<T>(ctx.GetPlace()), C);

      if ((N * sample_size) == 1) {
272 273
        // Only 1 element in normalization dimension,
        // we skip the batch norm calculation, let y = x.
274
        framework::TensorCopy(*x, ctx.GetPlace(), y);
275 276 277
        return;
      }

Q
QI JUN 已提交
278 279
      switch (data_layout) {
        case DataLayout::kNCHW: {
Q
Qiao Longfei 已提交
280 281 282 283 284 285 286 287 288 289 290 291
          ConstEigenArrayMap<T> x_arr(x->data<T>(), sample_size, N * C);
          for (int nc = 0; nc < N * C; ++nc) {
            saved_mean_e(nc % C) += x_arr.col(nc).sum();
          }
          saved_mean_e /= N * sample_size;
          for (int nc = 0; nc < N * C; ++nc) {
            saved_variance_e(nc % C) +=
                (x_arr.col(nc) - saved_mean_e(nc % C)).matrix().squaredNorm();
          }
          saved_variance_e /= N * sample_size;
          break;
        }
Q
QI JUN 已提交
292
        case DataLayout::kNHWC: {
Q
Qiao Longfei 已提交
293 294 295 296 297 298 299 300 301 302 303 304 305
          ConstEigenArrayMap<T> x_arr(x->data<T>(), C, N * sample_size);
          for (int i = 0; i < N * sample_size; ++i) {
            saved_mean_e += x_arr.col(i);
          }
          saved_mean_e /= N * sample_size;
          for (int i = 0; i < N * sample_size; ++i) {
            saved_variance_e +=
                (x_arr.col(i) - saved_mean_e) * (x_arr.col(i) - saved_mean_e);
          }
          saved_variance_e /= N * sample_size;
          break;
        }
        default:
Q
QI JUN 已提交
306
          PADDLE_THROW("Unknown storage order: %s", data_layout_str);
Q
Qiao Longfei 已提交
307 308 309 310 311 312 313 314 315 316
      }

      running_mean_arr =
          running_mean_arr * momentum + saved_mean_e * (1. - momentum);
      running_var_arr =
          running_var_arr * momentum + saved_variance_e * (1. - momentum);
    }

    // use SavedMean and SavedVariance to do normalize
    Eigen::Array<T, Eigen::Dynamic, 1> inv_std(C);
317
    if (global_stats) {
Q
Qiao Longfei 已提交
318 319 320 321 322 323 324 325 326 327 328
      ConstEigenVectorArrayMap<T> var_arr(
          ctx.Input<Tensor>("Variance")->data<T>(), C);
      inv_std = (var_arr + epsilon).sqrt().inverse();
    } else {
      EigenVectorArrayMap<T> saved_inv_std(
          ctx.Output<Tensor>("SavedVariance")->data<T>(), C);
      // inverse SavedVariance first, gradient will use it too.
      saved_inv_std = (saved_inv_std + epsilon).inverse().sqrt();
      inv_std = saved_inv_std;
    }
    ConstEigenVectorArrayMap<T> mean_arr(
329 330
        global_stats ? ctx.Input<Tensor>("Mean")->data<T>()
                     : ctx.Output<Tensor>("SavedMean")->data<T>(),
Q
Qiao Longfei 已提交
331 332 333 334 335 336 337 338 339 340 341 342 343
        C);

    //   ((x - est_mean) * (inv_var) * scale + bias
    //   formula transform ====>
    //   (x * inv_var * scale) + (bias - est_mean * inv_var * scale)
    const auto *scale = ctx.Input<Tensor>("Scale");
    const auto *bias = ctx.Input<Tensor>("Bias");
    ConstEigenVectorArrayMap<T> scale_arr(scale->data<T>(), C);
    ConstEigenVectorArrayMap<T> bias_arr(bias->data<T>(), C);
    Eigen::Array<T, Eigen::Dynamic, 1> new_scale = inv_std * scale_arr;
    Eigen::Array<T, Eigen::Dynamic, 1> new_bias =
        bias_arr - mean_arr * inv_std * scale_arr;

Q
QI JUN 已提交
344 345
    switch (data_layout) {
      case DataLayout::kNCHW: {
Q
Qiao Longfei 已提交
346 347 348 349 350 351 352 353
        EigenArrayMap<T> y_arr(y->mutable_data<T>(ctx.GetPlace()), sample_size,
                               N * C);
        ConstEigenArrayMap<T> x_arr(x->data<T>(), sample_size, N * C);
        for (int nc = 0; nc < N * C; ++nc) {
          y_arr.col(nc) = x_arr.col(nc) * new_scale(nc % C) + new_bias(nc % C);
        }
        break;
      }
Q
QI JUN 已提交
354
      case DataLayout::kNHWC: {
Q
Qiao Longfei 已提交
355 356 357 358 359 360 361 362 363
        EigenArrayMap<T>(y->mutable_data<T>(ctx.GetPlace()), C,
                         N * sample_size) =
            (ConstEigenArrayMap<T>(x->data<T>(), C, N * sample_size).colwise() *
             new_scale)
                .colwise() +
            new_bias;
        break;
      }
      default:
Q
QI JUN 已提交
364
        PADDLE_THROW("Unknown storage order: %d", data_layout);
Q
Qiao Longfei 已提交
365 366 367 368
    }
  }
};

Q
qingqing01 已提交
369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
void BatchNormGradOp::InferShape(framework::InferShapeContext *ctx) const {
  // check input
  PADDLE_ENFORCE(ctx->HasInput("X"));
  PADDLE_ENFORCE(ctx->HasInput("Scale"), "Input(scale) should not be null.");
  PADDLE_ENFORCE(ctx->HasInput(framework::GradVarName("Y")),
                 "Input(Y@GRAD) should not be null.");
  PADDLE_ENFORCE(ctx->HasInput("SavedMean"),
                 "Input(SavedMean) should not be null.");
  PADDLE_ENFORCE(ctx->HasInput("SavedVariance"),
                 "Input(SavedVariance) should not be null");

  // check output
  PADDLE_ENFORCE(ctx->HasOutput(framework::GradVarName("X")), "");
  if (ctx->HasOutput(framework::GradVarName("Scale"))) {
    PADDLE_ENFORCE(ctx->HasOutput(framework::GradVarName("Bias")),
                   "Output(Scale@GRAD) and Output(Bias@GRAD) should not be "
                   "null at same time");
  }
  const bool use_global_stats = ctx->Attrs().Get<bool>("use_global_stats");
  if (use_global_stats) {
    PADDLE_ENFORCE(!ctx->Attrs().Get<bool>("use_mkldnn"),
                   "Using global stats during training is not supported "
                   "in gradient op kernel of batch_norm_mkldnn_op now.");
  }
Q
Qiao Longfei 已提交
393

Q
qingqing01 已提交
394 395 396 397 398
  const auto x_dims = ctx->GetInputDim("X");
  const DataLayout data_layout = framework::StringToDataLayout(
      ctx->Attrs().Get<std::string>("data_layout"));
  const int C = (data_layout == DataLayout::kNCHW ? x_dims[1]
                                                  : x_dims[x_dims.size() - 1]);
Q
Qiao Longfei 已提交
399

Q
qingqing01 已提交
400 401 402 403
  ctx->SetOutputDim(framework::GradVarName("X"), x_dims);
  if (ctx->HasOutput(framework::GradVarName("Scale"))) {
    ctx->SetOutputDim(framework::GradVarName("Scale"), {C});
    ctx->SetOutputDim(framework::GradVarName("Bias"), {C});
Q
Qiao Longfei 已提交
404
  }
Q
qingqing01 已提交
405
}
Q
Qiao Longfei 已提交
406

Q
qingqing01 已提交
407 408 409 410 411 412 413 414 415 416 417 418 419 420 421
framework::OpKernelType BatchNormGradOp::GetExpectedKernelType(
    const framework::ExecutionContext &ctx) const {
  const auto *var = ctx.InputVar(framework::GradVarName("Y"));
  if (var == nullptr) {
    PADDLE_THROW("can't find Y@GRAD");
  }
  const Tensor *t = nullptr;
  if (var->IsType<Tensor>()) {
    t = &var->Get<Tensor>();
  } else if (var->IsType<LoDTensor>()) {
    t = &var->Get<LoDTensor>();
  }
  if (t == nullptr) {
    PADDLE_THROW("can't find Y@GRAD");
  }
422

Q
qingqing01 已提交
423 424 425
  // TODO(pzelazko-intel): enable MKLDNN layout when it's ready
  framework::LibraryType library = framework::LibraryType::kPlain;
  framework::DataLayout layout = framework::DataLayout::kAnyLayout;
426

427
#ifdef PADDLE_WITH_MKLDNN
Q
qingqing01 已提交
428 429 430 431 432
  if (library == framework::LibraryType::kPlain &&
      platform::CanMKLDNNBeUsed(ctx)) {
    library = framework::LibraryType::kMKLDNN;
    layout = framework::DataLayout::kMKLDNN;
  }
433
#endif
434

Q
qingqing01 已提交
435 436 437
  return framework::OpKernelType(ctx.Input<Tensor>("X")->type(), ctx.GetPlace(),
                                 layout, library);
}
Q
Qiao Longfei 已提交
438 439

template <typename T>
Q
QI JUN 已提交
440
class BatchNormGradKernel<platform::CPUDeviceContext, T>
Q
Qiao Longfei 已提交
441 442 443 444 445 446 447 448 449
    : public framework::OpKernel<T> {
 public:
  void Compute(const framework::ExecutionContext &ctx) const override {
    const auto *x = ctx.Input<Tensor>("X");
    const auto *d_y = ctx.Input<Tensor>(framework::GradVarName("Y"));
    const auto *scale = ctx.Input<Tensor>("Scale");
    const auto *saved_mean = ctx.Input<Tensor>("SavedMean");
    // SavedVariance have been reverted in forward operator
    const auto *saved_inv_variance = ctx.Input<Tensor>("SavedVariance");
Q
QI JUN 已提交
450
    const std::string data_layout_str = ctx.Attr<std::string>("data_layout");
451 452
    const bool use_global_stats = ctx.Attr<bool>("use_global_stats");
    const float epsilon = ctx.Attr<float>("epsilon");
Q
QI JUN 已提交
453 454
    const DataLayout data_layout =
        framework::StringToDataLayout(data_layout_str);
Q
Qiao Longfei 已提交
455 456 457 458

    // Get the size for each dimension.
    // NCHW [batch_size, in_channels, in_height, in_width]
    const auto &x_dims = x->dims();
459 460
    PADDLE_ENFORCE(x_dims.size() >= 2 && x_dims.size() <= 5,
                   "The Input dim size should be between 2 and 5");
Q
Qiao Longfei 已提交
461 462
    const int N = x_dims[0];
    const int C =
Q
QI JUN 已提交
463 464
        (data_layout == DataLayout::kNCHW ? x_dims[1]
                                          : x_dims[x_dims.size() - 1]);
Q
Qiao Longfei 已提交
465 466 467 468 469 470 471 472
    const int sample_size = x->numel() / N / C;

    // init output
    auto *d_x = ctx.Output<Tensor>(framework::GradVarName("X"));
    auto *d_scale = ctx.Output<Tensor>(framework::GradVarName("Scale"));
    auto *d_bias = ctx.Output<Tensor>(framework::GradVarName("Bias"));

    d_x->mutable_data<T>(ctx.GetPlace());
473 474 475 476 477 478 479 480

    const T *mean_data = saved_mean->data<T>();
    const T *inv_var_data = saved_inv_variance->data<T>();
    Tensor inv_var_tensor;
    if (use_global_stats) {
      const auto *running_mean = ctx.Input<Tensor>("Mean");
      const auto *running_variance = ctx.Input<Tensor>("Variance");
      mean_data = running_mean->data<T>();
Z
Zeng Jinle 已提交
481
      inv_var_tensor.Resize({C});
482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501
      T *running_inv_var_data = inv_var_tensor.mutable_data<T>(ctx.GetPlace());
      EigenVectorArrayMap<T> inv_var_tmp(running_inv_var_data, C);
      ConstEigenVectorArrayMap<T> var_arr(running_variance->data<T>(), C);

      inv_var_tmp = (var_arr + epsilon).sqrt().inverse().eval();
      inv_var_data = running_inv_var_data;
    }

    ConstEigenVectorArrayMap<T> scale_arr(scale->data<T>(), C);
    ConstEigenVectorArrayMap<T> mean_arr(mean_data, C);
    ConstEigenVectorArrayMap<T> inv_var_arr(inv_var_data, C);

    T *d_bias_data = nullptr;
    T *d_scale_data = nullptr;
    if (d_scale && d_bias) {
      d_scale->mutable_data<T>(ctx.GetPlace());
      d_bias->mutable_data<T>(ctx.GetPlace());
      d_bias_data = d_bias->mutable_data<T>(ctx.GetPlace());
      d_scale_data = d_scale->mutable_data<T>(ctx.GetPlace());
    }
Q
Qiao Longfei 已提交
502 503 504 505 506

    // d_bias = np.sum(d_y, axis=0)
    // d_scale = np.sum((X - mean) / inv_std * dy, axis=0)
    // d_x = (1. / N) * scale * inv_var * (N * d_y - np.sum(d_y, axis=0)
    //   - (X - mean) * inv_var * inv_var * np.sum(d_y * (X - mean), axis=0))
507 508
    EigenVectorArrayMap<T> d_bias_arr(d_bias_data, C);
    EigenVectorArrayMap<T> d_scale_arr(d_scale_data, C);
Q
Qiao Longfei 已提交
509

510 511 512 513
    if (d_scale && d_bias) {
      d_bias_arr.setZero();
      d_scale_arr.setZero();
    }
Q
Qiao Longfei 已提交
514

515 516
    if ((N * sample_size) == 1 && !use_global_stats) {
      framework::TensorCopy(*d_y, ctx.GetPlace(), d_x);
517 518 519
      return;
    }

520 521
    int scale_coefff = use_global_stats ? 1 : N * sample_size;
    const auto scale_inv_var_nhw = scale_arr * inv_var_arr / scale_coefff;
Q
Qiao Longfei 已提交
522

L
lvmengsi 已提交
523 524 525 526 527 528 529 530 531 532 533 534 535 536 537
    Tensor dy_sum;
    dy_sum.Resize({C});
    dy_sum.mutable_data<T>(ctx.GetPlace());
    EigenVectorArrayMap<T> dy_sum_arr(dy_sum.mutable_data<T>(ctx.GetPlace()),
                                      C);

    Tensor dy_mul_x_sub_mean_mul_invstd_sum;
    dy_mul_x_sub_mean_mul_invstd_sum.Resize({C});
    dy_mul_x_sub_mean_mul_invstd_sum.mutable_data<T>(ctx.GetPlace());
    EigenVectorArrayMap<T> dy_mul_x_sub_mean_mul_invstd_sum_arr(
        dy_mul_x_sub_mean_mul_invstd_sum.mutable_data<T>(ctx.GetPlace()), C);

    dy_sum_arr.setZero();
    dy_mul_x_sub_mean_mul_invstd_sum_arr.setZero();

Q
QI JUN 已提交
538 539
    switch (data_layout) {
      case DataLayout::kNCHW: {
Q
Qiao Longfei 已提交
540 541 542 543 544 545
        ConstEigenArrayMap<T> x_arr(x->data<T>(), sample_size, N * C);
        ConstEigenArrayMap<T> d_y_arr(d_y->data<T>(), sample_size, N * C);
        EigenArrayMap<T> d_x_arr(d_x->mutable_data<T>(ctx.GetPlace()),
                                 sample_size, N * C);
        d_x_arr.setZero();

L
lvmengsi 已提交
546 547 548 549 550 551 552 553
        for (int nc = 0; nc < N * C; ++nc) {
          int c = nc % C;
          dy_sum_arr(c) += d_y_arr.col(nc).sum();
          dy_mul_x_sub_mean_mul_invstd_sum_arr(c) +=
              ((x_arr.col(nc) - mean_arr(c)) * inv_var_arr(c) * d_y_arr.col(nc))
                  .sum();
        }

554
        if (d_scale && d_bias) {
L
lvmengsi 已提交
555 556
          d_bias_arr = dy_sum_arr;
          d_scale_arr = dy_mul_x_sub_mean_mul_invstd_sum_arr;
Q
Qiao Longfei 已提交
557
        }
L
lvmengsi 已提交
558

559 560 561 562 563
        if (!use_global_stats) {
          for (int nc = 0; nc < N * C; ++nc) {
            int c = nc % C;
            d_x_arr.col(nc) +=
                scale_inv_var_nhw(c) *
L
lvmengsi 已提交
564 565 566
                (d_y_arr.col(nc) * N * sample_size - dy_sum_arr(c) -
                 (x_arr.col(nc) - mean_arr[c]) *
                     dy_mul_x_sub_mean_mul_invstd_sum_arr(c) * inv_var_arr(c));
567 568 569 570 571 572
          }
        } else {
          for (int nc = 0; nc < N * C; ++nc) {
            int c = nc % C;
            d_x_arr.col(nc) += scale_inv_var_nhw(c) * d_y_arr.col(nc);
          }
Q
Qiao Longfei 已提交
573 574 575
        }
        break;
      }
Q
QI JUN 已提交
576
      case DataLayout::kNHWC: {
Q
Qiao Longfei 已提交
577 578 579 580 581 582
        ConstEigenArrayMap<T> x_arr(x->data<T>(), C, N * sample_size);
        ConstEigenArrayMap<T> d_y_arr(d_y->data<T>(), C, N * sample_size);
        EigenArrayMap<T> d_x_arr(d_x->mutable_data<T>(ctx.GetPlace()), C,
                                 N * sample_size);
        d_x_arr.setZero();

L
lvmengsi 已提交
583 584 585 586 587
        for (int nhw = 0; nhw < N * sample_size; ++nhw) {
          dy_sum_arr += d_y_arr.col(nhw);
          dy_mul_x_sub_mean_mul_invstd_sum_arr +=
              (x_arr.col(nhw) - mean_arr) * inv_var_arr * d_y_arr.col(nhw);
        }
588 589

        if (d_scale && d_bias) {
L
lvmengsi 已提交
590 591
          d_bias_arr = dy_sum_arr;
          d_scale_arr = dy_mul_x_sub_mean_mul_invstd_sum_arr;
592 593 594 595 596 597
        }

        if (!use_global_stats) {
          for (int nhw = 0; nhw < N * sample_size; ++nhw) {
            d_x_arr.col(nhw) +=
                scale_inv_var_nhw *
L
lvmengsi 已提交
598 599 600
                (d_y_arr.col(nhw) * N * sample_size - dy_sum_arr -
                 (x_arr.col(nhw) - mean_arr) *
                     dy_mul_x_sub_mean_mul_invstd_sum_arr * inv_var_arr);
601 602 603 604 605
          }
        } else {
          for (int nhw = 0; nhw < N * sample_size; ++nhw) {
            d_x_arr.col(nhw) += scale_inv_var_nhw * d_y_arr.col(nhw);
          }
Q
Qiao Longfei 已提交
606 607 608 609
        }
        break;
      }
      default:
Q
QI JUN 已提交
610
        PADDLE_THROW("Unknown storage order: %s", data_layout_str);
Q
Qiao Longfei 已提交
611 612 613 614
    }
  }
};

Q
qingqing01 已提交
615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630
std::unique_ptr<framework::OpDesc> BatchNormGradMaker::Apply() const {
  auto *op = new framework::OpDesc();
  op->SetType(GradOpType());
  op->SetInput("X", Input("X"));
  op->SetInput(framework::GradVarName("Y"), OutputGrad("Y"));

  op->SetInput("Scale", Input("Scale"));
  op->SetInput("Bias", Input("Bias"));
  op->SetInput("SavedMean", Output("SavedMean"));
  op->SetInput("SavedVariance", Output("SavedVariance"));

  // used when setting use_global_stats True during training
  if (boost::get<bool>(GetAttr("use_global_stats"))) {
    op->SetInput("Mean", Output("MeanOut"));
    op->SetInput("Variance", Output("VarianceOut"));
  }
631

Q
qingqing01 已提交
632
  op->SetAttrMap(Attrs());
Y
Yu Yang 已提交
633

Q
qingqing01 已提交
634 635 636
  op->SetOutput(framework::GradVarName("X"), InputGrad("X"));
  op->SetOutput(framework::GradVarName("Scale"), InputGrad("Scale"));
  op->SetOutput(framework::GradVarName("Bias"), InputGrad("Bias"));
Y
Yu Yang 已提交
637

Q
qingqing01 已提交
638 639
  return std::unique_ptr<framework::OpDesc>(op);
}
Y
Yu Yang 已提交
640

Q
Qiao Longfei 已提交
641 642 643 644
}  // namespace operators
}  // namespace paddle

namespace ops = paddle::operators;
Y
Yu Yang 已提交
645
REGISTER_OPERATOR(batch_norm, ops::BatchNormOp, ops::BatchNormOpMaker,
646 647
                  ops::BatchNormOpInferVarType, ops::BatchNormGradMaker);
REGISTER_OPERATOR(batch_norm_grad, ops::BatchNormGradOp);
Y
Yu Yang 已提交
648

Q
QI JUN 已提交
649
REGISTER_OP_CPU_KERNEL(
D
dzhwinter 已提交
650 651
    batch_norm, ops::BatchNormKernel<paddle::platform::CPUDeviceContext, float>,
    ops::BatchNormKernel<paddle::platform::CPUDeviceContext, double>);
Q
Qiao Longfei 已提交
652 653
REGISTER_OP_CPU_KERNEL(
    batch_norm_grad,
D
dzhwinter 已提交
654 655
    ops::BatchNormGradKernel<paddle::platform::CPUDeviceContext, float>,
    ops::BatchNormGradKernel<paddle::platform::CPUDeviceContext, double>);