batch_norm_op.cu 30.7 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. */

15
#include <algorithm>
Q
Qiao Longfei 已提交
16
#include <cfloat>
17 18 19
#include <string>
#include <vector>
#include "cub/cub.cuh"
S
Siddharth Goyal 已提交
20
#include "paddle/fluid/framework/data_layout.h"
21
#include "paddle/fluid/operators/batch_norm_op.h"
Y
Yi Wang 已提交
22 23
#include "paddle/fluid/operators/math/math_function.h"
#include "paddle/fluid/platform/cudnn_helper.h"
K
Kexin Zhao 已提交
24
#include "paddle/fluid/platform/float16.h"
Q
Qiao Longfei 已提交
25

26
DECLARE_bool(cudnn_batchnorm_spatial_persistent);
W
Wu Yi 已提交
27

Q
Qiao Longfei 已提交
28 29 30 31
namespace paddle {
namespace operators {

using Tensor = framework::Tensor;
Q
QI JUN 已提交
32
using DataLayout = framework::DataLayout;
Q
Qiao Longfei 已提交
33 34
template <typename T>
using CudnnDataType = platform::CudnnDataType<T>;
K
Kexin Zhao 已提交
35
template <typename T>
K
update  
Kexin Zhao 已提交
36
using BatchNormParamType = typename CudnnDataType<T>::BatchNormParamType;
Q
Qiao Longfei 已提交
37 38

template <typename T>
Q
QI JUN 已提交
39 40
class BatchNormKernel<platform::CUDADeviceContext, T>
    : public framework::OpKernel<T> {
Q
Qiao Longfei 已提交
41 42 43
 public:
  void Compute(const framework::ExecutionContext &ctx) const override {
    PADDLE_ENFORCE(platform::is_gpu_place(ctx.GetPlace()),
D
dzhwinter 已提交
44
                   "It must use CUDAPlace.");
Q
Qiao Longfei 已提交
45
    double epsilon = static_cast<double>(ctx.Attr<float>("epsilon"));
46
    float momentum = ctx.Attr<float>("momentum");
Q
Qiao Longfei 已提交
47
    const bool is_test = ctx.Attr<bool>("is_test");
48
    const bool use_global_stats = ctx.Attr<bool>("use_global_stats");
Q
QI JUN 已提交
49 50 51
    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
52 53 54 55 56

    // Get the size for each dimension.
    // NCHW [batch_size, in_channels, in_height, in_width]
    const auto *x = ctx.Input<Tensor>("X");
    const auto &x_dims = x->dims();
57 58
    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
59

60 61 62
    auto *y = ctx.Output<Tensor>("Y");
    y->mutable_data<T>(ctx.GetPlace());

63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91
    int N, C, H, W, D;
    ExtractNCWHD(x_dims, data_layout, &N, &C, &H, &W, &D);

    auto dtype = platform::CudnnDataType<T>::type;
    const bool fast_nhwc_batch_norm =
        is_test ||
        (dtype == CUDNN_DATA_HALF && FLAGS_cudnn_batchnorm_spatial_persistent);

    auto compute_format =
        fast_nhwc_batch_norm && data_layout == DataLayout::kNHWC
            ? DataLayout::kNHWC
            : DataLayout::kNCHW;

    Tensor transformed_x(x->type());
    Tensor transformed_y(y->type());
    if (data_layout == DataLayout::kNHWC &&
        compute_format == DataLayout::kNCHW && x_dims.size() > 2) {
      VLOG(3) << "Transform input tensor from NHWC to NCHW.";
      ResizeToChannelFirst<platform::CUDADeviceContext, T>(ctx, x,
                                                           &transformed_x);
      TransToChannelFirst<platform::CUDADeviceContext, T>(ctx, x,
                                                          &transformed_x);
      ResizeToChannelFirst<platform::CUDADeviceContext, T>(ctx, y,
                                                           &transformed_y);
    } else {
      transformed_x.ShareDataWith(*x);
      transformed_y.ShareDataWith(*y);
    }

Q
Qiao Longfei 已提交
92 93 94 95 96 97 98 99 100 101 102 103 104 105 106
    // ------------------- cudnn descriptors ---------------------
    cudnnTensorDescriptor_t data_desc_;
    cudnnTensorDescriptor_t bn_param_desc_;
    cudnnBatchNormMode_t mode_;

    CUDNN_ENFORCE(platform::dynload::cudnnCreateTensorDescriptor(&data_desc_));
    CUDNN_ENFORCE(
        platform::dynload::cudnnCreateTensorDescriptor(&bn_param_desc_));

    if (epsilon <= CUDNN_BN_MIN_EPSILON - FLT_EPSILON) {
      LOG(ERROR) << "Provided epsilon is smaller than "
                 << "CUDNN_BN_MIN_EPSILON. Setting it to "
                 << "CUDNN_BN_MIN_EPSILON instead.";
    }
    epsilon = std::max(epsilon, CUDNN_BN_MIN_EPSILON);
107
#if CUDNN_VERSION_MIN(7, 0, 0)
W
Wu Yi 已提交
108 109 110 111 112
    if (FLAGS_cudnn_batchnorm_spatial_persistent) {
      mode_ = CUDNN_BATCHNORM_SPATIAL_PERSISTENT;
    } else {
      mode_ = CUDNN_BATCHNORM_SPATIAL;
    }
113
#else
Q
Qiao Longfei 已提交
114
    mode_ = CUDNN_BATCHNORM_SPATIAL;
115
#endif
Q
Qiao Longfei 已提交
116

M
minqiyang 已提交
117
    VLOG(3) << "Setting descriptors.";
Q
Qiao Longfei 已提交
118 119
    std::vector<int> dims;
    std::vector<int> strides;
120
    if (compute_format == DataLayout::kNCHW) {
Q
Qiao Longfei 已提交
121 122 123 124 125 126 127 128 129
      dims = {N, C, H, W, D};
      strides = {C * H * W * D, H * W * D, W * D, D, 1};
    } else {
      dims = {N, C, H, W, D};
      strides = {H * W * D * C, 1, W * D * C, D * C, C};
    }
    CUDNN_ENFORCE(platform::dynload::cudnnSetTensorNdDescriptor(
        data_desc_, CudnnDataType<T>::type,
        x_dims.size() > 3 ? x_dims.size() : 4, dims.data(), strides.data()));
K
Kexin Zhao 已提交
130
    // Note: PERSISTENT not implemented for inference
Q
Qiao Longfei 已提交
131
    CUDNN_ENFORCE(platform::dynload::cudnnDeriveBNTensorDescriptor(
K
Kexin Zhao 已提交
132
        bn_param_desc_, data_desc_, is_test ? CUDNN_BATCHNORM_SPATIAL : mode_));
Q
Qiao Longfei 已提交
133 134 135 136

    const auto *scale = ctx.Input<Tensor>("Scale");
    const auto *bias = ctx.Input<Tensor>("Bias");

Q
QI JUN 已提交
137
    auto &dev_ctx = ctx.template device_context<platform::CUDADeviceContext>();
Q
Qiao Longfei 已提交
138

Q
QI JUN 已提交
139
    auto handle = dev_ctx.cudnn_handle();
Q
Qiao Longfei 已提交
140 141

    // Now, depending on whether we are running test or not, we have two paths.
142
    if (is_test || use_global_stats) {
Q
Qiao Longfei 已提交
143 144 145 146 147 148 149 150 151 152 153 154 155
      // only when test we use input to do computation.
      const auto *est_mean = ctx.Input<Tensor>("Mean");
      const auto *est_var = ctx.Input<Tensor>("Variance");
      // Run inference mode.
      PADDLE_ENFORCE_EQ(est_mean->dims().size(), 1UL);
      PADDLE_ENFORCE_EQ(est_var->dims().size(), 1UL);
      PADDLE_ENFORCE_EQ(est_mean->dims()[0], C);
      PADDLE_ENFORCE_EQ(est_var->dims()[0], C);

      CUDNN_ENFORCE(platform::dynload::cudnnBatchNormalizationForwardInference(
          handle,
          // Note: PERSISTENT not implemented for inference
          CUDNN_BATCHNORM_SPATIAL, CudnnDataType<T>::kOne(),
156 157 158
          CudnnDataType<T>::kZero(), data_desc_,
          transformed_x.template data<T>(), data_desc_,
          transformed_y.template mutable_data<T>(ctx.GetPlace()),
K
update  
Kexin Zhao 已提交
159 160 161 162
          bn_param_desc_, scale->template data<BatchNormParamType<T>>(),
          bias->template data<BatchNormParamType<T>>(),
          est_mean->template data<BatchNormParamType<T>>(),
          est_var->template data<BatchNormParamType<T>>(), epsilon));
Q
Qiao Longfei 已提交
163
    } else {
164 165 166 167 168 169 170 171 172
      // if MomentumTensor is set, use MomentumTensor value, momentum
      // is only used in this training branch
      if (ctx.HasInput("MomentumTensor")) {
        const auto *mom_tensor = ctx.Input<Tensor>("MomentumTensor");
        Tensor mom_cpu;
        TensorCopySync(*mom_tensor, platform::CPUPlace(), &mom_cpu);
        momentum = mom_cpu.data<float>()[0];
      }

Q
Qiao Longfei 已提交
173 174 175
      // Run training mode.
      // obtain running mean and running inv var, and see if we need to
      // initialize them.
D
Dang Qingqing 已提交
176 177 178 179 180 181 182 183 184 185 186 187 188 189 190

      auto *mean_out = ctx.Output<Tensor>("MeanOut");
      auto *variance_out = ctx.Output<Tensor>("VarianceOut");
      mean_out->mutable_data<BatchNormParamType<T>>(ctx.GetPlace());
      variance_out->mutable_data<BatchNormParamType<T>>(ctx.GetPlace());

      auto *saved_mean = ctx.Output<Tensor>("SavedMean");
      auto *saved_variance = ctx.Output<Tensor>("SavedVariance");
      saved_mean->mutable_data<BatchNormParamType<T>>(ctx.GetPlace());
      saved_variance->mutable_data<BatchNormParamType<T>>(ctx.GetPlace());
      math::SetConstant<platform::CUDADeviceContext, BatchNormParamType<T>>
          functor;
      functor(dev_ctx, saved_mean, static_cast<BatchNormParamType<T>>(0));
      functor(dev_ctx, saved_variance, static_cast<BatchNormParamType<T>>(0));

191
      if ((N * H * W * D) == 1) {
192 193
        // Only 1 element in normalization dimension,
        // skip the batch norm calculation, let y = x.
194
        framework::TensorCopy(*x, ctx.GetPlace(), y);
195 196 197
      } else {
        double this_factor = 1. - momentum;

198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284
        bool called = false;
#if CUDNN_VERSION_MIN(7, 4, 1)
        if (compute_format == DataLayout::kNHWC) {
          called = true;
          size_t workspace_size = 0;
          size_t reserve_space_size = 0;
          void *reserve_space_ptr = nullptr;
          void *workspace_ptr = nullptr;
          Tensor workspace_tensor;
          // Create reserve space and workspace for batch norm.
          // Create tensor for each batchnorm op, it will be used in the
          // backward. Thus this tensor shouldn't be temp.
          auto *reserve_space = ctx.Output<Tensor>("ReserveSpace");
          PADDLE_ENFORCE_NOT_NULL(
              reserve_space,
              platform::errors::NotFound(
                  "The argument ReserveSpace of batch_norm op is not found."));

          // --------------- cudnn batchnorm workspace ---------------
          CUDNN_ENFORCE(
              platform::dynload::
                  cudnnGetBatchNormalizationForwardTrainingExWorkspaceSize(
                      /*handle=*/handle,
                      /*mode=*/mode_,
                      /*bnIps=*/CUDNN_BATCHNORM_OPS_BN,
                      /*xDesc=*/data_desc_,
                      /*zDesc=*/nullptr,
                      /*yDesc=*/data_desc_,
                      /*bnScaleBiasMeanVarDesc=*/bn_param_desc_,
                      /*activationDesc=*/nullptr,
                      /*sizeInBytes=*/&workspace_size));

          // -------------- cudnn batchnorm reserve space --------------
          CUDNN_ENFORCE(
              platform::dynload::
                  cudnnGetBatchNormalizationTrainingExReserveSpaceSize(
                      /*handle=*/handle,
                      /*mode=*/mode_,
                      /*bnOps=*/CUDNN_BATCHNORM_OPS_BN,
                      /*activationDesc=*/nullptr,
                      /*xDesc=*/data_desc_,
                      /*sizeInBytes=*/&reserve_space_size));

          reserve_space_ptr = reserve_space->mutable_data(
              ctx.GetPlace(), transformed_x.type(), reserve_space_size);
          workspace_ptr = workspace_tensor.mutable_data(
              ctx.GetPlace(), transformed_x.type(), workspace_size);
          CUDNN_ENFORCE(
              platform::dynload::cudnnBatchNormalizationForwardTrainingEx(
                  handle, mode_, CUDNN_BATCHNORM_OPS_BN,
                  CudnnDataType<T>::kOne(), CudnnDataType<T>::kZero(),
                  data_desc_, transformed_x.template data<T>(), nullptr,
                  nullptr, data_desc_, transformed_y.template data<T>(),
                  bn_param_desc_, scale->template data<BatchNormParamType<T>>(),
                  bias->template data<BatchNormParamType<T>>(), this_factor,
                  mean_out->template mutable_data<BatchNormParamType<T>>(
                      ctx.GetPlace()),
                  variance_out->template mutable_data<BatchNormParamType<T>>(
                      ctx.GetPlace()),
                  epsilon,
                  saved_mean->template mutable_data<BatchNormParamType<T>>(
                      ctx.GetPlace()),
                  saved_variance->template mutable_data<BatchNormParamType<T>>(
                      ctx.GetPlace()),
                  nullptr, workspace_ptr, workspace_size, reserve_space_ptr,
                  reserve_space_size));
        }
#endif
        if (!called) {
          CUDNN_ENFORCE(
              platform::dynload::cudnnBatchNormalizationForwardTraining(
                  handle, mode_, CudnnDataType<T>::kOne(),
                  CudnnDataType<T>::kZero(), data_desc_,
                  transformed_x.template data<T>(), data_desc_,
                  transformed_y.template mutable_data<T>(ctx.GetPlace()),
                  bn_param_desc_, scale->template data<BatchNormParamType<T>>(),
                  bias->template data<BatchNormParamType<T>>(), this_factor,
                  mean_out->template mutable_data<BatchNormParamType<T>>(
                      ctx.GetPlace()),
                  variance_out->template mutable_data<BatchNormParamType<T>>(
                      ctx.GetPlace()),
                  epsilon,
                  saved_mean->template mutable_data<BatchNormParamType<T>>(
                      ctx.GetPlace()),
                  saved_variance->template mutable_data<BatchNormParamType<T>>(
                      ctx.GetPlace())));
        }
285
      }
Q
Qiao Longfei 已提交
286 287
    }

288 289 290 291 292 293
    if (data_layout == DataLayout::kNHWC &&
        compute_format == DataLayout::kNCHW && x_dims.size() > 2) {
      VLOG(3) << "Transform batchnorm output from NCHW to NHWC";
      TransToChannelLast<paddle::platform::CUDADeviceContext, T>(
          ctx, &transformed_y, y);
    }
Q
Qiao Longfei 已提交
294 295 296 297 298 299 300
    // clean when exit.
    CUDNN_ENFORCE(platform::dynload::cudnnDestroyTensorDescriptor(data_desc_));
    CUDNN_ENFORCE(
        platform::dynload::cudnnDestroyTensorDescriptor(bn_param_desc_));
  }
};

301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336
template <typename T, int BlockDim, framework::DataLayout layout>
static __global__ void KeBNBackwardScaleBias(
    const T *dy, const T *x, const BatchNormParamType<T> *mean,
    const BatchNormParamType<T> *variance, const double epsilon, const int N,
    const int C, const int HxW, BatchNormParamType<T> *dscale,
    BatchNormParamType<T> *dbias) {
  const int outer_size = C;
  const int inner_size = N * HxW;
  typedef cub::BlockReduce<BatchNormParamType<T>, BlockDim> BlockReduce;
  __shared__ typename BlockReduce::TempStorage ds_storage;
  __shared__ typename BlockReduce::TempStorage db_storage;

  for (int i = blockIdx.x; i < outer_size; i += gridDim.x) {
    BatchNormParamType<T> ds_sum = static_cast<BatchNormParamType<T>>(0);
    BatchNormParamType<T> db_sum = static_cast<BatchNormParamType<T>>(0);

    BatchNormParamType<T> inv_var_i = 1.0 / sqrt(variance[i] + epsilon);
    BatchNormParamType<T> mean_i = mean[i];
    for (int j = threadIdx.x; j < inner_size; j += blockDim.x) {
      const int index = layout == framework::DataLayout::kNCHW
                            ? (j / HxW * C + i) * HxW + j % HxW
                            : j * outer_size + i;
      ds_sum += static_cast<BatchNormParamType<T>>(dy[index]) *
                (static_cast<BatchNormParamType<T>>(x[index]) - mean_i);
      db_sum += static_cast<BatchNormParamType<T>>(dy[index]);
    }
    ds_sum = BlockReduce(ds_storage).Reduce(ds_sum, cub::Sum());
    db_sum = BlockReduce(db_storage).Reduce(db_sum, cub::Sum());
    if (threadIdx.x == 0) {
      dscale[i] = ds_sum * inv_var_i;
      dbias[i] = db_sum;
    }
    __syncthreads();
  }
}

Q
qingqing01 已提交
337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352
template <typename T, framework::DataLayout layout>
static __global__ void KeBNBackwardData(const T *dy,
                                        const BatchNormParamType<T> *scale,
                                        const BatchNormParamType<T> *variance,
                                        const double epsilon, const int C,
                                        const int HxW, const int num, T *dx) {
  int gid = blockIdx.x * blockDim.x + threadIdx.x;
  int stride = blockDim.x * gridDim.x;
  for (int i = gid; i < num; i += stride) {
    const int c = layout == framework::DataLayout::kNCHW ? i / HxW % C : i % C;
    BatchNormParamType<T> inv_var = 1.0 / sqrt(variance[c] + epsilon);
    dx[i] = static_cast<T>(static_cast<BatchNormParamType<T>>(dy[i]) *
                           scale[c] * inv_var);
  }
}

L
lvmengsi 已提交
353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409
template <typename T, int BlockDim, framework::DataLayout layout>
static __global__ void BNBackwardData(const T *dy,
                                      const BatchNormParamType<T> *scale,
                                      const BatchNormParamType<T> *mean,
                                      const T *x,
                                      const BatchNormParamType<T> *variance,
                                      const int C, const int N, const int HxW,
                                      T *dx) {
  const int outer_size = C;
  const int inner_size = N * HxW;
  typedef cub::BlockReduce<BatchNormParamType<T>, BlockDim> BlockReduce;
  __shared__ typename BlockReduce::TempStorage dy_storage;
  __shared__ typename BlockReduce::TempStorage dy_x_sub_mean_storage;
  __shared__ BatchNormParamType<T> dy_sum_val;
  __shared__ BatchNormParamType<T> dy_x_sub_mean_sum_val;

  for (int i = blockIdx.x; i < outer_size; i += gridDim.x) {
    BatchNormParamType<T> inv_var_i = variance[i];
    BatchNormParamType<T> mean_i = mean[i];
    BatchNormParamType<T> dy_sum = static_cast<BatchNormParamType<T>>(0);
    BatchNormParamType<T> dy_x_sub_mean_sum =
        static_cast<BatchNormParamType<T>>(0);
    for (int j = threadIdx.x; j < inner_size; j += blockDim.x) {
      const int index = layout == framework::DataLayout::kNCHW
                            ? (j / HxW * C + i) * HxW + j % HxW
                            : j * outer_size + i;
      BatchNormParamType<T> dy_i =
          static_cast<BatchNormParamType<T>>(dy[index]);
      dy_sum += dy_i;
      dy_x_sub_mean_sum +=
          dy_i * (static_cast<BatchNormParamType<T>>(x[index]) - mean_i);
    }

    dy_sum = BlockReduce(dy_storage).Reduce(dy_sum, cub::Sum());
    dy_x_sub_mean_sum = BlockReduce(dy_x_sub_mean_storage)
                            .Reduce(dy_x_sub_mean_sum, cub::Sum());

    if (threadIdx.x == 0) {
      dy_sum_val = dy_sum;
      dy_x_sub_mean_sum_val = dy_x_sub_mean_sum;
    }
    __syncthreads();

    for (int j = threadIdx.x; j < inner_size; j += blockDim.x) {
      const int index = layout == framework::DataLayout::kNCHW
                            ? (j / HxW * C + i) * HxW + j % HxW
                            : j * outer_size + i;
      dx[index] =
          (static_cast<BatchNormParamType<T>>(dy[index]) -
           dy_sum_val / static_cast<BatchNormParamType<T>>(inner_size) -
           (static_cast<BatchNormParamType<T>>(x[index]) - mean_i) *
               dy_x_sub_mean_sum_val * inv_var_i * inv_var_i / inner_size) *
          scale[i] * inv_var_i;
    }
  }
}

Q
Qiao Longfei 已提交
410
template <typename T>
Q
QI JUN 已提交
411
class BatchNormGradKernel<platform::CUDADeviceContext, T>
Q
Qiao Longfei 已提交
412 413 414 415
    : public framework::OpKernel<T> {
 public:
  void Compute(const framework::ExecutionContext &ctx) const override {
    PADDLE_ENFORCE(platform::is_gpu_place(ctx.GetPlace()),
D
dzhwinter 已提交
416
                   "It must use CUDAPlace.");
Q
Qiao Longfei 已提交
417
    double epsilon = static_cast<double>(ctx.Attr<float>("epsilon"));
Q
QI JUN 已提交
418
    const std::string data_layout_str = ctx.Attr<std::string>("data_layout");
419 420
    const bool use_global_stats = ctx.Attr<bool>("use_global_stats");

Q
QI JUN 已提交
421 422
    const DataLayout data_layout =
        framework::StringToDataLayout(data_layout_str);
Q
Qiao Longfei 已提交
423 424 425 426 427 428
    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 &x_dims = x->dims();

429 430
    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
431
    int N, C, H, W, D;
Q
QI JUN 已提交
432
    ExtractNCWHD(x_dims, data_layout, &N, &C, &H, &W, &D);
Q
Qiao Longfei 已提交
433

434 435 436 437 438 439
    // 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());
440 441 442
    if (d_scale && d_bias) {
      d_scale->mutable_data<BatchNormParamType<T>>(ctx.GetPlace());
      d_bias->mutable_data<BatchNormParamType<T>>(ctx.GetPlace());
443
    }
Q
Qiao Longfei 已提交
444 445 446
    PADDLE_ENFORCE_EQ(scale->dims().size(), 1UL);
    PADDLE_ENFORCE_EQ(scale->dims()[0], C);

447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478
    auto dtype = platform::CudnnDataType<T>::type;
    const auto *reserve_space = ctx.Input<Tensor>("ReserveSpace");
    const bool fast_nhwc_batch_norm =
        dtype == CUDNN_DATA_HALF && FLAGS_cudnn_batchnorm_spatial_persistent &&
        reserve_space != nullptr;
    auto compute_format =
        fast_nhwc_batch_norm && data_layout == DataLayout::kNHWC
            ? DataLayout::kNHWC
            : DataLayout::kNCHW;

    Tensor transformed_x(x->type());
    Tensor transformed_d_y(d_y->type());
    Tensor transformed_d_x(d_x->type());
    if (data_layout == DataLayout::kNHWC &&
        compute_format == DataLayout::kNCHW) {
      VLOG(3) << "Transform input tensor from NHWC to NCHW.";
      ResizeToChannelFirst<platform::CUDADeviceContext, T>(ctx, x,
                                                           &transformed_x);
      TransToChannelFirst<platform::CUDADeviceContext, T>(ctx, x,
                                                          &transformed_x);
      ResizeToChannelFirst<platform::CUDADeviceContext, T>(ctx, d_y,
                                                           &transformed_d_y);
      TransToChannelFirst<platform::CUDADeviceContext, T>(ctx, d_y,
                                                          &transformed_d_y);
      ResizeToChannelFirst<platform::CUDADeviceContext, T>(ctx, d_x,
                                                           &transformed_d_x);
    } else {
      transformed_x.ShareDataWith(*x);
      transformed_d_y.ShareDataWith(*d_y);
      transformed_d_x.ShareDataWith(*d_x);
    }

Z
zchen0211 已提交
479 480
    std::vector<int> dims;
    std::vector<int> strides;
481
    if (compute_format == DataLayout::kNCHW) {
Z
zchen0211 已提交
482 483 484 485 486 487
      dims = {N, C, H, W, D};
      strides = {C * H * W * D, H * W * D, W * D, D, 1};
    } else {
      dims = {N, C, H, W, D};
      strides = {H * W * C * D, 1, W * D * C, D * C, C};
    }
Q
Qiao Longfei 已提交
488

489
    auto &dev_ctx = ctx.template device_context<platform::CUDADeviceContext>();
490
    const int num = transformed_x.numel();
L
lvmengsi 已提交
491 492 493 494 495 496
    const int block = 512;
    int max_threads = dev_ctx.GetMaxPhysicalThreadCount();
    const int max_blocks = std::max(max_threads / block, 1);
    int grid1 = (num + block - 1) / block;
    int grid2 = std::min(C, max_blocks);

497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521
    if (!use_global_stats) {
      if ((N * H * W * D) == 1) {
        framework::TensorCopy(*d_y, ctx.GetPlace(), d_x);
        math::SetConstant<platform::CUDADeviceContext, BatchNormParamType<T>>
            functor;
        functor(dev_ctx, d_scale, static_cast<BatchNormParamType<T>>(0));
        functor(dev_ctx, d_bias, static_cast<BatchNormParamType<T>>(0));
        return;
      }

      // ------------------- cudnn descriptors ---------------------
      cudnnTensorDescriptor_t data_desc_;
      cudnnTensorDescriptor_t bn_param_desc_;
      cudnnBatchNormMode_t mode_;

      CUDNN_ENFORCE(
          platform::dynload::cudnnCreateTensorDescriptor(&data_desc_));
      CUDNN_ENFORCE(
          platform::dynload::cudnnCreateTensorDescriptor(&bn_param_desc_));
      if (epsilon <= CUDNN_BN_MIN_EPSILON - FLT_EPSILON) {
        LOG(ERROR) << "Provided epsilon is smaller than "
                   << "CUDNN_BN_MIN_EPSILON. Setting it to "
                   << "CUDNN_BN_MIN_EPSILON instead.";
      }
      epsilon = std::max(epsilon, CUDNN_BN_MIN_EPSILON);
522
#if CUDNN_VERSION_MIN(7, 0, 0)
W
Wu Yi 已提交
523 524 525 526 527
      if (FLAGS_cudnn_batchnorm_spatial_persistent) {
        mode_ = CUDNN_BATCHNORM_SPATIAL_PERSISTENT;
      } else {
        mode_ = CUDNN_BATCHNORM_SPATIAL;
      }
528
#else
529
      mode_ = CUDNN_BATCHNORM_SPATIAL;
530
#endif
531 532 533 534 535 536 537 538 539

      CUDNN_ENFORCE(platform::dynload::cudnnSetTensorNdDescriptor(
          data_desc_, CudnnDataType<T>::type,
          x_dims.size() > 3 ? x_dims.size() : 4, dims.data(), strides.data()));
      CUDNN_ENFORCE(platform::dynload::cudnnDeriveBNTensorDescriptor(
          bn_param_desc_, data_desc_, mode_));

      const auto *saved_mean = ctx.Input<Tensor>("SavedMean");
      const auto *saved_var = ctx.Input<Tensor>("SavedVariance");
L
lvmengsi 已提交
540
      const auto *saved_mean_data =
541
          saved_mean->template data<BatchNormParamType<T>>();
L
lvmengsi 已提交
542
      const auto *saved_var_data =
543 544
          saved_var->template data<BatchNormParamType<T>>();

L
lvmengsi 已提交
545
      if (d_scale && d_bias) {
546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632
        bool called = false;
#if CUDNN_VERSION_MIN(7, 4, 1)
        if (compute_format == DataLayout::kNHWC) {
          called = true;
          size_t workspace_size = 0;
          void *workspace_ptr = nullptr;
          Tensor workspace_tensor;
          auto reserve_space_size = reserve_space->memory_size();
          // --------------- cudnn batchnorm workspace ---------------
          CUDNN_ENFORCE(platform::dynload::
                            cudnnGetBatchNormalizationBackwardExWorkspaceSize(
                                /*handle=*/dev_ctx.cudnn_handle(),
                                /*mode=*/mode_,
                                /*bnIps=*/CUDNN_BATCHNORM_OPS_BN,
                                /*xDesc=*/data_desc_,
                                /*yDesc=*/data_desc_,
                                /*dyDesc=*/data_desc_,
                                /*dzDesc=*/nullptr,
                                /*dxDesc=*/data_desc_,
                                /*bnScaleBiasMeanVarDesc=*/bn_param_desc_,
                                /*activationDesc=*/nullptr,
                                /*sizeInBytes=*/&workspace_size));

          workspace_ptr = workspace_tensor.mutable_data(
              ctx.GetPlace(), transformed_x.type(), workspace_size);

          CUDNN_ENFORCE(platform::dynload::cudnnBatchNormalizationBackwardEx(
              /*handle=*/dev_ctx.cudnn_handle(),
              /*mode=*/mode_,
              /*bnOps=*/CUDNN_BATCHNORM_OPS_BN,
              /*alphaDataDiff=*/CudnnDataType<T>::kOne(),
              /*betaDataDiff=*/CudnnDataType<T>::kZero(),
              /*alphaParamDiff=*/CudnnDataType<T>::kOne(),
              /*betaParamDiff=*/CudnnDataType<T>::kZero(),
              /*xDesc=*/data_desc_,
              /*xData=*/transformed_x.template data<T>(),
              /*yDesc=*/nullptr,
              /*yData=*/nullptr,
              /*dyDesc=*/data_desc_,
              /*dyData=*/transformed_d_y.template data<T>(),
              /*dzDesc=*/nullptr,
              /*dzData=*/nullptr,
              /*dxDesc=*/data_desc_,
              /*dxData=*/transformed_d_x.template mutable_data<T>(
                  ctx.GetPlace()),
              /*dBnScaleBiasDesc=*/bn_param_desc_,
              /*bnScaleData=*/scale->template data<BatchNormParamType<T>>(),
              /*bnBiasData=*/nullptr,
              /*dBnScaleData=*/d_scale
                  ->template mutable_data<BatchNormParamType<T>>(
                      ctx.GetPlace()),
              /*dBnBiasData=*/d_bias
                  ->template mutable_data<BatchNormParamType<T>>(
                      ctx.GetPlace()),
              /*epsilon=*/epsilon,
              /*savedMean=*/saved_mean_data,
              /*savedInvVariance=*/saved_var_data,
              /*activationDesc=*/nullptr,
              /*workspace=*/workspace_ptr,
              /*workSpaceSizeInBytes=*/workspace_size,
              /*reserveSpace=*/const_cast<T *>(
                  reserve_space->template data<T>()),
              /*reserveSpaceSizeInBytes=*/reserve_space_size));
        }
#endif
        if (!called) {
          CUDNN_ENFORCE(platform::dynload::cudnnBatchNormalizationBackward(
              dev_ctx.cudnn_handle(), mode_, CudnnDataType<T>::kOne(),
              CudnnDataType<T>::kZero(), CudnnDataType<T>::kOne(),
              CudnnDataType<T>::kZero(), data_desc_,
              transformed_x.template data<T>(), data_desc_,
              transformed_d_y.template data<T>(), data_desc_,
              transformed_d_x.template mutable_data<T>(ctx.GetPlace()),
              bn_param_desc_, scale->template data<BatchNormParamType<T>>(),
              d_scale->template mutable_data<BatchNormParamType<T>>(
                  ctx.GetPlace()),
              d_bias->template mutable_data<BatchNormParamType<T>>(
                  ctx.GetPlace()),
              epsilon, saved_mean_data, saved_var_data));
        }

        if (data_layout == DataLayout::kNHWC &&
            compute_format == DataLayout::kNCHW) {
          VLOG(3) << "Transform batchnorm output from NCHW to NHWC";
          TransToChannelLast<paddle::platform::CUDADeviceContext, T>(
              ctx, &transformed_d_x, d_x);
        }
L
lvmengsi 已提交
633
      } else {
634
        if (compute_format == DataLayout::kNCHW) {
L
lvmengsi 已提交
635 636 637 638 639 640 641 642 643
          if (d_x) {
            BNBackwardData<T, block, framework::DataLayout::kNCHW><<<
                grid2, block, 0, dev_ctx.stream()>>>(
                d_y->data<T>(), scale->data<BatchNormParamType<T>>(),
                saved_mean_data, x->data<T>(), saved_var_data, C, N, H * W * D,
                d_x->data<T>());
          }
        } else {
          if (d_x) {
L
Lv Mengsi 已提交
644
            BNBackwardData<T, block, framework::DataLayout::kNHWC><<<
L
lvmengsi 已提交
645 646 647 648 649 650 651
                grid2, block, 0, dev_ctx.stream()>>>(
                d_y->data<T>(), scale->data<BatchNormParamType<T>>(),
                saved_mean_data, x->data<T>(), saved_var_data, C, N, H * W * D,
                d_x->data<T>());
          }
        }
      }
652 653 654 655 656 657 658 659 660 661 662 663 664 665 666

      // clean when exit.
      CUDNN_ENFORCE(
          platform::dynload::cudnnDestroyTensorDescriptor(data_desc_));
      CUDNN_ENFORCE(
          platform::dynload::cudnnDestroyTensorDescriptor(bn_param_desc_));
    } else {
      const auto *running_mean = ctx.Input<Tensor>("Mean");
      const auto *running_var = ctx.Input<Tensor>("Variance");

      const auto *running_mean_data =
          running_mean->template data<BatchNormParamType<T>>();
      const auto *running_var_data =
          running_var->template data<BatchNormParamType<T>>();

667
      if (compute_format == DataLayout::kNCHW) {
668 669 670 671 672 673 674 675 676 677
        if (d_x) {
          KeBNBackwardData<T, framework::DataLayout::kNCHW><<<
              grid1, block, 0, dev_ctx.stream()>>>(
              d_y->data<T>(), scale->data<BatchNormParamType<T>>(),
              running_var_data, epsilon, C, H * W, num, d_x->data<T>());
        }
        if (d_scale && d_bias) {
          KeBNBackwardScaleBias<T, block, framework::DataLayout::kNCHW><<<
              grid2, block, 0, dev_ctx.stream()>>>(
              d_y->data<T>(), x->data<T>(), running_mean_data, running_var_data,
Q
qingqing01 已提交
678
              epsilon, N, C, H * W * D, d_scale->data<BatchNormParamType<T>>(),
679 680 681 682 683 684 685 686 687 688
              d_bias->data<BatchNormParamType<T>>());
        }
      } else {
        if (d_x) {
          KeBNBackwardData<T, framework::DataLayout::kNHWC><<<
              grid1, block, 0, dev_ctx.stream()>>>(
              d_y->data<T>(), scale->data<BatchNormParamType<T>>(),
              running_var_data, epsilon, C, H * W, num, d_x->data<T>());
        }
        if (d_scale && d_bias) {
Q
qingqing01 已提交
689
          KeBNBackwardScaleBias<T, block, framework::DataLayout::kNHWC><<<
690 691
              grid2, block, 0, dev_ctx.stream()>>>(
              d_y->data<T>(), x->data<T>(), running_mean_data, running_var_data,
Q
qingqing01 已提交
692
              epsilon, N, C, H * W * D, d_scale->data<BatchNormParamType<T>>(),
693 694 695 696
              d_bias->data<BatchNormParamType<T>>());
        }
      }
    }
Q
Qiao Longfei 已提交
697 698 699 700 701 702 703
  }
};

}  // namespace operators
}  // namespace paddle

namespace ops = paddle::operators;
K
Kexin Zhao 已提交
704
namespace plat = paddle::platform;
Q
QI JUN 已提交
705
REGISTER_OP_CUDA_KERNEL(
K
Kexin Zhao 已提交
706
    batch_norm, ops::BatchNormKernel<plat::CUDADeviceContext, float>,
D
dzhwinter 已提交
707
    ops::BatchNormKernel<plat::CUDADeviceContext, double>,
K
Kexin Zhao 已提交
708
    ops::BatchNormKernel<plat::CUDADeviceContext, plat::float16>);
Q
QI JUN 已提交
709
REGISTER_OP_CUDA_KERNEL(
D
dzhwinter 已提交
710
    batch_norm_grad, ops::BatchNormGradKernel<plat::CUDADeviceContext, float>,
C
chengduo 已提交
711 712
    ops::BatchNormGradKernel<plat::CUDADeviceContext, double>,
    ops::BatchNormGradKernel<plat::CUDADeviceContext, plat::float16>);