reduce_op.h 29.6 KB
Newer Older
1
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
G
guosheng 已提交
2

L
Luo Tao 已提交
3 4 5
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
G
guosheng 已提交
6

L
Luo Tao 已提交
7
    http://www.apache.org/licenses/LICENSE-2.0
G
guosheng 已提交
8

L
Luo Tao 已提交
9 10 11 12 13
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. */
G
guosheng 已提交
14 15 16

#pragma once

17
#include <algorithm>
18
#include <set>
19
#include <string>
W
whs 已提交
20
#include <vector>
21
#include "paddle/fluid/framework/data_type_transform.h"
22
#include "paddle/fluid/framework/tensor_util.h"
23
#include "paddle/fluid/operators/cast_op.h"
24
#include "paddle/fluid/operators/math/math_function.h"
W
Wu Yi 已提交
25
#include "paddle/fluid/operators/reduce_ops/reduce_op_function.h"
26 27 28

// only can include the headers in paddle/pten/api dirs
#include "paddle/pten/api/lib/utils/tensor_utils.h"
29
#include "paddle/pten/kernels/cpu/reduce.h"
30

31
#if defined(__HIPCC__) || defined(__NVCC__)
32
#include "paddle/pten/kernels/gpu/reduce.h"
33
#endif
G
guosheng 已提交
34 35 36 37

namespace paddle {
namespace operators {

38 39
#define HANDLE_DIM(NDIM, RDIM)                                            \
  if (ndim == NDIM && rdim == RDIM) {                                     \
40 41
    paddle::operators::ReduceFunctor<DeviceContext, OutT, NDIM, RDIM,     \
                                     Functor>(                            \
42 43
        context.template device_context<DeviceContext>(), *input, output, \
        dims, keep_dim);                                                  \
W
whs 已提交
44 45
  }

46
using Tensor = framework::Tensor;
47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71
using DDim = framework::DDim;

inline void GetShuffledDim(const DDim& src_dims, DDim* dst_dims,
                           const std::vector<int>& reduced_dims,
                           std::vector<int>* perm_axis) {
  // check if it's a reduced dim
  std::vector<bool> src_dims_check(src_dims.size(), false);
  size_t src_size = src_dims.size();
  size_t reduce_size = reduced_dims.size();
  for (size_t i = 0; i < reduce_size; ++i) {
    dst_dims->at(src_size - reduce_size + i) = src_dims[reduced_dims[i]];
    (*perm_axis)[src_size - reduce_size + i] = reduced_dims[i];
    src_dims_check[reduced_dims[i]] = true;
  }

  size_t offset = 0;
  for (size_t i = 0; i < src_dims_check.size(); ++i) {
    bool is_reduced = src_dims_check[i];
    if (!is_reduced) {
      (*perm_axis)[offset] = i;
      dst_dims->at(offset++) = src_dims[i];
    }
  }
}

72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92
static inline std::vector<int> GetReduceDim(const std::vector<int>& dims,
                                            int dim_size, bool reduce_all) {
  std::vector<int> reduce_dims;
  if (reduce_all) {
    reduce_dims.resize(dim_size);
    int reduce_size = reduce_dims.size();
    for (int i = 0; i < reduce_size; ++i) {
      reduce_dims[i] = i;
    }
  } else {
    for (auto e : dims) {
      PADDLE_ENFORCE_LT(e, dim_size,
                        paddle::platform::errors::InvalidArgument(
                            "ReduceOp: invalid axis, when x_dims is %d, "
                            "axis[i] should less than x_dims, but got %d.",
                            dim_size, e));
      reduce_dims.push_back(e >= 0 ? e : e + dim_size);
    }
  }
  return reduce_dims;
}
93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134
template <typename DeviceContext, typename OutT>
void GetShuffledInput(const framework::ExecutionContext& context,
                      const Tensor* input, Tensor* shuffled_input,
                      const std::vector<int>& dims) {
  DDim shuffled_dims(input->dims());
  std::vector<int> perm_axis(input->dims().size());
  GetShuffledDim(input->dims(), &shuffled_dims, dims, &perm_axis);

  shuffled_input->Resize(shuffled_dims);
  shuffled_input->mutable_data<OutT>(context.GetPlace());

  math::TransposeNormal<DeviceContext, OutT> trans;
  trans(context.template device_context<DeviceContext>(), *input,
        shuffled_input, perm_axis);
}

inline void GetOriginDimFromShuffled(const DDim& src_dim,
                                     const std::vector<int>& dims,
                                     std::vector<int>* origin_dim) {
  DDim shuffled_dims(src_dim);
  size_t n = src_dim.size();
  std::vector<int> perm_axis(n);
  GetShuffledDim(src_dim, &shuffled_dims, dims, &perm_axis);
  for (size_t i = 0; i < n; ++i) {
    (*origin_dim)[perm_axis[i]] = i;
  }
}

template <typename DeviceContext, typename OutT, typename Functor>
void HandleLargeDim(const framework::ExecutionContext& context,
                    const Tensor* input, Tensor* output,
                    const std::vector<int>& dims, bool keep_dim) {
  //  shuffle the reduced dim to the end
  Tensor shuffled_input;
  GetShuffledInput<DeviceContext, OutT>(context, input, &shuffled_input, dims);

  // transpose to 2D tensor whose shape is {unreduced, reduced}.
  const int64_t unreduced = output->numel();
  const int64_t reduced = shuffled_input.numel() / unreduced;
  shuffled_input.Resize({unreduced, reduced});
  DDim output_dim = output->dims();
  output->Resize({unreduced});
135
  paddle::operators::ReduceFunctor<DeviceContext, OutT, 2, 1, Functor>(
136 137 138 139 140 141 142 143 144 145
      context.template device_context<DeviceContext>(), shuffled_input, output,
      {1}, keep_dim);
  output->Resize(output_dim);
}

template <typename DeviceContext, typename T, typename Functor>
void HandleLargeDimGrad(const framework::ExecutionContext& context,
                        const framework::Tensor* x,
                        const framework::Tensor* out,
                        const framework::Tensor* dout, framework::Tensor* dx,
146
                        Functor functor, const std::vector<int>& dims) {
147 148 149 150 151 152 153 154 155 156 157 158 159
  const int64_t unreduced = out->numel();
  const int64_t reduced = x->numel() / unreduced;
  DDim out_dim(out->dims());
  DDim x_dim(x->dims());
  // transpose and reshape X
  Tensor shuffled_x;
  GetShuffledInput<DeviceContext, T>(context, x, &shuffled_x, dims);
  DDim shuffled_dim = shuffled_x.dims();
  shuffled_x.Resize({unreduced, reduced});
  // reshape dX {unreduced, reduced}
  dx->Resize({unreduced, reduced});
  ReduceGradFunctor<DeviceContext, T, 2, Functor>(
      context.template device_context<DeviceContext>(), shuffled_x, *out, *dout,
160
      dx, functor, {1});
161 162 163 164 165 166 167 168 169 170 171
  // transpose dX
  std::vector<int> origin_axis(x_dim.size());
  GetOriginDimFromShuffled(x_dim, dims, &origin_axis);
  Tensor dx_tmp;
  framework::TensorCopy(*dx, context.GetPlace(), &dx_tmp);
  dx_tmp.Resize(shuffled_dim);
  dx->Resize(x_dim);
  math::TransposeNormal<DeviceContext, T> trans;
  trans(context.template device_context<DeviceContext>(), dx_tmp, dx,
        origin_axis);
}
172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206

template <typename DeviceContext, typename T, typename Functor>
struct ReduceKernelFunctor {
  const Tensor* input;
  Tensor* output;
  std::vector<int> dims;
  bool keep_dim;
  bool reduce_all;
  const framework::ExecutionContext& context;
  ReduceKernelFunctor(const Tensor* input, Tensor* output,
                      const std::vector<int>& dims, bool keep_dim,
                      bool reduce_all,
                      const framework::ExecutionContext& context)
      : input(input),
        output(output),
        dims(dims),
        keep_dim(keep_dim),
        reduce_all(reduce_all),
        context(context) {}

  template <typename OutT>
  void apply() const {
    output->mutable_data<OutT>(context.GetPlace());
    if (reduce_all) {
      // Flatten and reduce 1-D tensor
      auto x = EigenVector<OutT>::Flatten(*input);
      auto out = EigenScalar<OutT>::From(*output);
      auto& place =
          *context.template device_context<DeviceContext>().eigen_device();
      auto reduce_dim = Eigen::array<int, 1>({{0}});
      Functor functor;
      functor(place, &x, &out, reduce_dim);
    } else {
      int ndim = input->dims().size();
      int rdim = dims.size();
207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227
      if (ndim > 6) {
        HandleLargeDim<DeviceContext, OutT, Functor>(context, input, output,
                                                     dims, keep_dim);
      } else {
        HANDLE_DIM(6, 5);
        HANDLE_DIM(6, 4);
        HANDLE_DIM(6, 3);
        HANDLE_DIM(6, 2);
        HANDLE_DIM(6, 1);
        HANDLE_DIM(5, 4);
        HANDLE_DIM(5, 3);
        HANDLE_DIM(5, 2);
        HANDLE_DIM(5, 1);
        HANDLE_DIM(4, 3);
        HANDLE_DIM(4, 2);
        HANDLE_DIM(4, 1);
        HANDLE_DIM(3, 2);
        HANDLE_DIM(3, 1);
        HANDLE_DIM(2, 1);
        HANDLE_DIM(1, 1);
      }
228 229 230
    }
  }
};
Q
QI JUN 已提交
231
template <typename DeviceContext, typename T, typename Functor>
Y
Yu Yang 已提交
232
class ReduceKernel : public framework::OpKernel<T> {
233 234 235 236 237 238 239 240
 public:
  void Compute(const framework::ExecutionContext& context) const override {
    bool reduce_all = context.Attr<bool>("reduce_all");
    auto* output = context.Output<Tensor>("Out");
    auto dims = context.Attr<std::vector<int>>("dim");
    bool keep_dim = context.Attr<bool>("keep_dim");
    int out_dtype = context.Attr<int>("out_dtype");
    framework::proto::VarType::Type cast_out_dtype;
241
    auto* input = context.Input<Tensor>("X");
242

243 244
    if (out_dtype < 0) {
      cast_out_dtype =
245
          static_cast<framework::proto::VarType::Type>(input->type());
246 247 248
    } else {
      cast_out_dtype = static_cast<framework::proto::VarType::Type>(out_dtype);
    }
249 250 251 252 253 254 255 256 257 258 259 260

    auto& dev_ctx = context.device_context<DeviceContext>();
    output->mutable_data(
        dev_ctx.GetPlace(),
        static_cast<framework::proto::VarType::Type>(cast_out_dtype));

    auto pt_x = paddle::experimental::MakePtenDenseTensor(*input);
    auto pt_out = paddle::experimental::MakePtenDenseTensor(*output);

    std::vector<int64_t> tmp_dims(dims.begin(), dims.end());

    // call new kernel
W
Wilber 已提交
261 262 263 264 265
    pten::Reduce<typename framework::ConvertToPtenContext<DeviceContext>::TYPE,
                 T, Functor>(
        static_cast<const typename framework::ConvertToPtenContext<
            DeviceContext>::TYPE&>(dev_ctx),
        *pt_x.get(), reduce_all, tmp_dims, keep_dim,
266
        pten::TransToPtenDataType(cast_out_dtype), pt_out.get());
267 268 269 270
  }
};
template <typename DeviceContext, typename OutT, typename Functor>
class BoolReduceKernel : public framework::OpKernel<OutT> {
G
guosheng 已提交
271 272
 public:
  void Compute(const framework::ExecutionContext& context) const override {
273
    bool reduce_all = context.Attr<bool>("reduce_all");
274 275
    auto* input = context.Input<Tensor>("X");
    auto* output = context.Output<Tensor>("Out");
276
    output->mutable_data<OutT>(context.GetPlace());
277 278 279 280

    auto dims = context.Attr<std::vector<int>>("dim");
    bool keep_dim = context.Attr<bool>("keep_dim");

281 282 283 284 285 286 287 288 289 290 291 292
    // The dims has full dim, set the reduce_all is True
    const auto& input_dim_size = context.Input<Tensor>("X")->dims().size();
    std::set<int> dims_set(dims.begin(), dims.end());
    bool full_dim = true;
    for (auto i = 0; i < input_dim_size; i++) {
      if (dims_set.find(i) == dims_set.end()) {
        full_dim = false;
        break;
      }
    }
    reduce_all = (reduce_all || full_dim);

293 294
    if (reduce_all) {
      // Flatten and reduce 1-D tensor
295 296
      auto x = EigenVector<OutT>::Flatten(*input);
      auto out = EigenScalar<OutT>::From(*output);
297 298 299 300
      auto& place =
          *context.template device_context<DeviceContext>().eigen_device();
      auto reduce_dim = Eigen::array<int, 1>({{0}});
      Functor functor;
301
      functor(place, &x, &out, reduce_dim);
302
    } else {
303 304
      int ndim = input->dims().size();
      int rdim = dims.size();
305
      // comments for accelerating compiling temporarily.
306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326
      if (ndim > 6) {
        HandleLargeDim<DeviceContext, OutT, Functor>(context, input, output,
                                                     dims, keep_dim);
      } else {
        HANDLE_DIM(6, 5);
        HANDLE_DIM(6, 4);
        HANDLE_DIM(6, 3);
        HANDLE_DIM(6, 2);
        HANDLE_DIM(6, 1);
        HANDLE_DIM(5, 4);
        HANDLE_DIM(5, 3);
        HANDLE_DIM(5, 2);
        HANDLE_DIM(5, 1);
        HANDLE_DIM(4, 3);
        HANDLE_DIM(4, 2);
        HANDLE_DIM(4, 1);
        HANDLE_DIM(3, 2);
        HANDLE_DIM(3, 1);
        HANDLE_DIM(2, 1);
        HANDLE_DIM(1, 1);
      }
G
guosheng 已提交
327 328 329
    }
  }
};
330

331 332 333 334 335
template <typename DeviceContext, typename T, typename Functor>
void LaunchReduceGradKernel(const framework::ExecutionContext& context,
                            const framework::Tensor* input0,
                            const framework::Tensor* input1,
                            const framework::Tensor* input2,
336
                            paddle::framework::Tensor* output, Functor functor,
337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355
                            const std::vector<int>& dims,
                            bool reduce_all = false) {
  if (reduce_all) {
    auto x = EigenVector<T>::Flatten(*input0);
    auto x_reduce = EigenVector<T>::Flatten(*input1);
    auto x_reduce_grad = EigenVector<T>::Flatten(*input2);
    auto x_grad = EigenVector<T>::Flatten(*output);
    auto& place =
        *context.template device_context<DeviceContext>().eigen_device();
    auto broadcast_dim =
        Eigen::array<int, 1>({{static_cast<int>(input0->numel())}});
    functor(place, &x, &x_reduce, &x_grad, &x_reduce_grad, broadcast_dim,
            broadcast_dim[0]);
  } else {
    int rank = input0->dims().size();
    switch (rank) {
      case 1:
        ReduceGradFunctor<DeviceContext, T, 1, Functor>(
            context.template device_context<DeviceContext>(), *input0, *input1,
356
            *input2, output, functor, dims);
357 358 359 360
        break;
      case 2:
        ReduceGradFunctor<DeviceContext, T, 2, Functor>(
            context.template device_context<DeviceContext>(), *input0, *input1,
361
            *input2, output, functor, dims);
362 363 364 365
        break;
      case 3:
        ReduceGradFunctor<DeviceContext, T, 3, Functor>(
            context.template device_context<DeviceContext>(), *input0, *input1,
366
            *input2, output, functor, dims);
367 368 369 370
        break;
      case 4:
        ReduceGradFunctor<DeviceContext, T, 4, Functor>(
            context.template device_context<DeviceContext>(), *input0, *input1,
371
            *input2, output, functor, dims);
372 373 374 375
        break;
      case 5:
        ReduceGradFunctor<DeviceContext, T, 5, Functor>(
            context.template device_context<DeviceContext>(), *input0, *input1,
376
            *input2, output, functor, dims);
377 378 379 380
        break;
      case 6:
        ReduceGradFunctor<DeviceContext, T, 6, Functor>(
            context.template device_context<DeviceContext>(), *input0, *input1,
381
            *input2, output, functor, dims);
382 383
        break;
      default:
384 385
        HandleLargeDimGrad<DeviceContext, T, Functor>(
            context, input0, input1, input2, output, functor, dims);
386 387 388 389 390
        break;
    }
  }
}

391 392
template <typename DeviceContext, typename T, typename Functor,
          bool kNoNeedBufferX = false, bool kNoNeedBufferY = false>
Y
Yu Yang 已提交
393
class ReduceGradKernel : public framework::OpKernel<T> {
G
guosheng 已提交
394
 public:
395 396
  void ComputeFromInput(const Tensor* input2,
                        const framework::ExecutionContext& context) const {
397
    bool reduce_all = context.Attr<bool>("reduce_all");
398 399 400
    auto dims = context.Attr<std::vector<int>>("dim");
    auto* input0 = context.Input<Tensor>("X");
    auto* input1 = context.Input<Tensor>("Out");
401

402 403 404
    auto* output = context.Output<Tensor>(framework::GradVarName("X"));
    output->mutable_data<T>(context.GetPlace());

405 406 407 408 409 410 411 412 413 414 415
    // The dims has full dim, set the reduce_all is True
    const auto& input_dim_size = context.Input<Tensor>("X")->dims().size();
    std::set<int> dims_set(dims.begin(), dims.end());
    bool full_dim = true;
    for (auto i = 0; i < input_dim_size; i++) {
      if (dims_set.find(i) == dims_set.end()) {
        full_dim = false;
        break;
      }
    }
    reduce_all = (reduce_all || full_dim);
416 417 418 419 420 421 422 423 424 425 426
    // NOTE: EigenTensor::From() uses tensor->data()
    // if op has NoNeedBufferVarsInferer, the corresponding kNoNeedBufferX or
    // kNoNeedBufferY should set true
    // and use fake var that has same dims.
    if (kNoNeedBufferX) {
      input0 = output;
    }
    if (kNoNeedBufferY) {
      input1 = input2;
    }

427 428
    const std::vector<int> const_dims = dims;

L
lvmengsi 已提交
429 430 431
    // NOTE(dengkaipeng): Out is unnecessary in some reduce kernel and
    // not be set as Input in grad Maker, use Out_grad to replace here
    if (!input1) input1 = input2;
432 433 434 435
    Functor functor;
    LaunchReduceGradKernel<DeviceContext, T, Functor>(context, input0, input1,
                                                      input2, output, functor,
                                                      const_dims, reduce_all);
G
guosheng 已提交
436
  }
437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456

  void Compute(const framework::ExecutionContext& context) const override {
    int in_dtype = context.Attr<int>("in_dtype");
    if (in_dtype >= 0) {
      Tensor tmp_tensor;
      auto* pre_input = context.Input<Tensor>(framework::GradVarName("Out"));
      auto in_kernel_type =
          framework::OpKernelType(pre_input->type(), context.GetPlace());
      auto out_kernel_type = framework::OpKernelType(
          static_cast<framework::proto::VarType::Type>(in_dtype),
          context.GetPlace());
      framework::TransDataType(in_kernel_type, out_kernel_type, *pre_input,
                               &tmp_tensor);
      ComputeFromInput(&tmp_tensor, context);

    } else {
      auto* input2 = context.Input<Tensor>(framework::GradVarName("Out"));
      ComputeFromInput(input2, context);
    }
  }
457
};
G
guosheng 已提交
458

459 460 461
class ReduceOp : public framework::OperatorWithKernel {
 public:
  using framework::OperatorWithKernel::OperatorWithKernel;
G
guosheng 已提交
462

463
  void InferShape(framework::InferShapeContext* ctx) const override {
464 465
    OP_INOUT_CHECK(ctx->HasInput("X"), "Input", "X", "ReduceOp");
    OP_INOUT_CHECK(ctx->HasOutput("Out"), "Output", "Out", "ReduceOp");
466 467 468
    auto x_dims = ctx->GetInputDim("X");
    auto x_rank = x_dims.size();
    auto dims = ctx->Attrs().Get<std::vector<int>>("dim");
469 470 471 472 473 474
    PADDLE_ENFORCE_GT(dims.size(), 0,
                      platform::errors::InvalidArgument(
                          "The input dim dimensions of ReduceOp "
                          "should be greater than 0. But received the dim "
                          "dimesions of Reduce = %d.",
                          dims.size()));
475

476
    for (size_t i = 0; i < dims.size(); ++i) {
477
      PADDLE_ENFORCE_LT(dims[i], x_rank,
478 479 480 481 482
                        platform::errors::InvalidArgument(
                            "The reduce dim index %d should be in the "
                            "range [-dimension(X), dimension(X)] "
                            "which dimesion = %d. But received dim index = %d.",
                            i, x_rank, dims[i]));
483 484 485 486 487 488
      PADDLE_ENFORCE_GE(dims[i], -x_rank,
                        platform::errors::InvalidArgument(
                            "The reduce dim index %d should be in the "
                            "range [-dimension(X), dimension(X)] "
                            "which dimesion = %d. But received dim index = %d.",
                            i, x_rank, dims[i]));
489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514
      if (dims[i] < 0) dims[i] = x_rank + dims[i];
    }
    sort(dims.begin(), dims.end());
    bool reduce_all = ctx->Attrs().Get<bool>("reduce_all");
    bool keep_dim = ctx->Attrs().Get<bool>("keep_dim");
    if (reduce_all) {
      if (keep_dim)
        ctx->SetOutputDim(
            "Out", framework::make_ddim(std::vector<int64_t>(x_rank, 1)));
      else
        ctx->SetOutputDim("Out", {1});
    } else {
      auto dims_vector = vectorize(x_dims);
      if (keep_dim) {
        for (size_t i = 0; i < dims.size(); ++i) {
          dims_vector[dims[i]] = 1;
        }
      } else {
        const int kDelFlag = -2;
        for (size_t i = 0; i < dims.size(); ++i) {
          dims_vector[dims[i]] = kDelFlag;
        }
        dims_vector.erase(
            remove(dims_vector.begin(), dims_vector.end(), kDelFlag),
            dims_vector.end());
      }
515 516 517
      if (!keep_dim && dims_vector.size() == 0) {
        dims_vector.push_back(1);
      }
518 519
      auto out_dims = framework::make_ddim(dims_vector);
      ctx->SetOutputDim("Out", out_dims);
520
      if (dims.size() > 0 && dims[0] != 0) {
521 522 523 524 525
        // Only pass LoD when not reducing on the first dim.
        ctx->ShareLoD("X", /*->*/ "Out");
      }
    }
  }
526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543

  framework::OpKernelType GetExpectedKernelType(
      const framework::ExecutionContext& ctx) const override {
    // choose cudnn kernel if the runtime supported.
    auto input_data_type = OperatorWithKernel::IndicateVarDataType(ctx, "X");

    if (ctx.Input<paddle::framework::LoDTensor>("X")->dims().size() > 5)
      return framework::OpKernelType(input_data_type, ctx.GetPlace());

#ifdef PADDLE_WITH_MKLDNN
    if (this->CanMKLDNNBeUsed(ctx, input_data_type)) {
      return framework::OpKernelType(input_data_type, ctx.GetPlace(),
                                     framework::DataLayout::kMKLDNN,
                                     framework::LibraryType::kMKLDNN);
    }
#endif

    if (input_data_type == framework::proto::VarType::FP16) {
F
furnace 已提交
544 545 546
      PADDLE_ENFORCE_EQ(platform::is_gpu_place(ctx.GetPlace()) ||
                            platform::is_npu_place(ctx.GetPlace()),
                        true,
547
                        platform::errors::InvalidArgument(
F
furnace 已提交
548
                            "float16 can only be used on GPU or NPU place"));
549 550 551
    }
    return framework::OpKernelType(input_data_type, ctx.GetPlace());
  }
552 553 554

  framework::KernelSignature GetExpectedPtenKernelArgs(
      const framework::ExecutionContext& ctx) const override {
555
    bool reduce_all = ctx.Attr<bool>("reduce_all");
556 557
    if (Type() == "reduce_sum") {
      if (ctx.InputVar("X")->IsType<framework::LoDTensor>()) {
558 559 560 561
        if (!reduce_all) {
          return framework::KernelSignature(
              "sum", {"X"}, {"dim", "keep_dim", "out_dtype"}, {"Out"});
        }
562
        return framework::KernelSignature(
563
            "sum_raw", {"X"}, {"dim", "keep_dim", "reduce_all", "out_dtype"},
564 565 566 567 568
            {"Out"});
      }
    }
    if (Type() == "reduce_mean") {
      if (ctx.InputVar("X")->IsType<framework::LoDTensor>()) {
569 570 571 572
        if (!reduce_all) {
          return framework::KernelSignature("mean", {"X"}, {"dim", "keep_dim"},
                                            {"Out"});
        }
573
        return framework::KernelSignature(
574
            "mean_raw", {"X"}, {"dim", "keep_dim", "reduce_all"}, {"Out"});
575 576 577 578 579
      }
    }
    // TODO(chentianyu03): support other cases after selected rows added
    return framework::KernelSignature("reduce.unregistered", {}, {}, {});
  }
580 581
};

G
Guo Sheng 已提交
582 583 584 585 586 587 588 589 590 591 592 593 594
class ReduceOpUseInputPlace : public ReduceOp {
 public:
  using ReduceOp::ReduceOp;

 protected:
  framework::OpKernelType GetExpectedKernelType(
      const framework::ExecutionContext& ctx) const override {
    framework::OpKernelType kt = OperatorWithKernel::GetExpectedKernelType(ctx);
    kt.place_ = ctx.Input<framework::LoDTensor>("X")->place();
    return kt;
  }
};

595 596 597
class ReduceGradOp : public framework::OperatorWithKernel {
 public:
  using framework::OperatorWithKernel::OperatorWithKernel;
W
whs 已提交
598

599
  void InferShape(framework::InferShapeContext* ctx) const override {
600 601 602
    OP_INOUT_CHECK(ctx->HasInput("X"), "Input", "X", "ReduceOp");
    OP_INOUT_CHECK(ctx->HasInput(framework::GradVarName("Out")), "Input",
                   "Out@GRAD", "ReduceOp");
603 604 605
    auto x_dims = ctx->GetInputDim("X");
    auto x_rank = x_dims.size();
    auto dims = ctx->Attrs().Get<std::vector<int>>("dim");
W
whs 已提交
606
    for (size_t i = 0; i < dims.size(); ++i) {
607
      PADDLE_ENFORCE_LT(dims[i], x_rank,
608 609 610 611 612
                        platform::errors::InvalidArgument(
                            "The reduce dim index %d should be in the "
                            "range [-dimension(X), dimension(X)], "
                            "which dimesion = %d. But received dim index = %d.",
                            i, x_rank, dims[i]));
W
whs 已提交
613
      if (dims[i] < 0) dims[i] = x_rank + dims[i];
614 615 616 617 618 619
    }
    sort(dims.begin(), dims.end());
    auto x_grad_name = framework::GradVarName("X");
    if (ctx->HasOutput(x_grad_name)) {
      ctx->SetOutputDim(x_grad_name, x_dims);
      ctx->ShareLoD("X", /*->*/ x_grad_name);
W
whs 已提交
620
    }
621
  }
622 623 624 625

 protected:
  framework::OpKernelType GetExpectedKernelType(
      const framework::ExecutionContext& ctx) const override {
J
jakpiase 已提交
626 627 628 629 630
    int in_dtype = ctx.Attr<int>("in_dtype");
    auto input_data_type =
        (in_dtype >= 0) ? static_cast<framework::proto::VarType::Type>(in_dtype)
                        : OperatorWithKernel::IndicateVarDataType(
                              ctx, framework::GradVarName("Out"));
631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647
#ifdef PADDLE_WITH_MKLDNN
    auto CanMKLDNNReduceGradBeUsed = [&]() {
      auto dx_dims = ctx.Input<Tensor>("X")->dims();

      if (dx_dims.size() > 5) return false;  // max 5D tensor is supported

      return true;
    };
    if (this->CanMKLDNNBeUsed(ctx, input_data_type) &&
        CanMKLDNNReduceGradBeUsed()) {
      return framework::OpKernelType(input_data_type, ctx.GetPlace(),
                                     framework::DataLayout::kMKLDNN,
                                     framework::LibraryType::kMKLDNN);
    }
#endif

    return framework::OpKernelType(input_data_type, ctx.GetPlace());
648
  }
649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672
};

class ReduceOpMaker : public framework::OpProtoAndCheckerMaker {
 public:
  void Make() final {
    AddInput("X",
             "(Tensor) The input tensor. Tensors with rank at most 6 are "
             "supported.");
    AddOutput("Out", "(Tensor) The result tensor.");
    AddAttr<std::vector<int>>(
        "dim",
        "(list<int>, default {0}) The dimensions to reduce. "
        "Must be in the range [-rank(input), rank(input)). "
        "If `dim[i] < 0`, the dims[i] to reduce is `rank + dims[i]`. "
        "Note that reducing on the first dim will make the LoD info lost.")
        .SetDefault({0});
    AddAttr<bool>("keep_dim",
                  "(bool, default false) "
                  "If true, retain the reduced dimension with length 1.")
        .SetDefault(false);
    AddAttr<bool>("reduce_all",
                  "(bool, default false) "
                  "If true, output a scalar reduced along all dimensions.")
        .SetDefault(false);
673 674 675 676 677 678 679 680 681 682
    AddAttr<int>("in_dtype",
                 "(int, default -1)"
                 "The dtype of input, default value is -1, the user could not "
                 "set this value.")
        .SetDefault(-1);
    AddAttr<int>(
        "out_dtype",
        "(int, default -1)"
        "The dtype of output, default value is -1, the dtype is same as intput")
        .SetDefault(-1);
683 684
    AddAttr<bool>("use_mkldnn",
                  "(bool, default false) Only used in mkldnn kernel")
685 686
        .SetDefault(false)
        .AsExtra();
687 688
    AddComment(string::Sprintf(R"DOC(
%s Operator.
W
whs 已提交
689

690 691 692
This operator computes the %s of input tensor along the given dimension.
The result tensor has 1 fewer dimension than the input unless keep_dim is true.
If reduce_all is true, just reduce along all dimensions and output a scalar.
W
whs 已提交
693

694 695
)DOC",
                               GetOpType(), GetName()));
G
guosheng 已提交
696
  }
697 698 699 700

 protected:
  virtual std::string GetName() const = 0;
  virtual std::string GetOpType() const = 0;
G
guosheng 已提交
701 702
};

703
#if defined(__HIPCC__) || defined(__NVCC__)
704 705
template <typename T, template <typename> class ReduceOp,
          template <typename, typename> class TransformOp>
706 707 708 709 710 711 712 713 714
class ReduceCudaKernel : public framework::OpKernel<T> {
 public:
  void Compute(const framework::ExecutionContext& context) const override {
    bool reduce_all = context.Attr<bool>("reduce_all");
    const Tensor* input = context.Input<Tensor>("X");
    Tensor* output = context.Output<Tensor>("Out");
    auto out_dtype = context.Attr<int>("out_dtype");
    std::vector<int> dims = context.Attr<std::vector<int>>("dim");

715 716
    auto& dev_ctx = context.cuda_device_context();

717
    if (out_dtype >= 0) {
718 719 720
      output->mutable_data(
          dev_ctx.GetPlace(),
          static_cast<framework::proto::VarType::Type>(out_dtype));
721
    } else {
722 723 724
      output->mutable_data(
          dev_ctx.GetPlace(),
          static_cast<framework::proto::VarType::Type>(input->type()));
725
    }
726 727 728 729 730 731 732 733 734 735 736

    auto pt_x = paddle::experimental::MakePtenDenseTensor(*input);
    auto pt_out = paddle::experimental::MakePtenDenseTensor(*output);
    std::vector<int64_t> dims_int64{dims.begin(), dims.end()};

    auto pt_out_dtype = pten::TransToPtenDataType(
        static_cast<framework::proto::VarType::Type>(out_dtype));

    pten::Reduce<T, ReduceOp, TransformOp>(dev_ctx, *pt_x.get(), reduce_all,
                                           dims_int64, false, pt_out_dtype,
                                           pt_out.get());
737 738 739 740
  }
};
#endif

G
guosheng 已提交
741 742
}  // namespace operators
}  // namespace paddle
743

744 745
namespace ops = paddle::operators;

H
hong 已提交
746 747 748 749 750 751 752 753 754 755 756 757 758 759
#define REGISTER_REDUCE_OP(op_name)                                           \
  class __##op_name##Maker__ : public ops::ReduceOpMaker {                    \
   protected:                                                                 \
    virtual std::string GetName() const { return #op_name; }                  \
    virtual std::string GetOpType() const { return "Reduce " #op_name; }      \
  };                                                                          \
  REGISTER_OPERATOR(                                                          \
      op_name, ops::ReduceOp, __##op_name##Maker__,                           \
      paddle::framework::DefaultGradOpMaker<paddle::framework::OpDesc, true>, \
      paddle::framework::DefaultGradOpMaker<paddle::imperative::OpBase,       \
                                            true>);                           \
  REGISTER_OPERATOR(op_name##_grad, ops::ReduceGradOp)

#define REGISTER_REDUCE_OP_WITHOUT_GRAD(op_name, ...)                    \
760 761 762 763 764
  class __##op_name##Maker__ : public ops::ReduceOpMaker {               \
   protected:                                                            \
    virtual std::string GetName() const { return #op_name; }             \
    virtual std::string GetOpType() const { return "Reduce " #op_name; } \
  };                                                                     \
H
hong 已提交
765 766 767 768
  REGISTER_OPERATOR(                                                     \
      op_name, ops::ReduceOp##__VA_ARGS__, __##op_name##Maker__,         \
      paddle::framework::EmptyGradOpMaker<paddle::framework::OpDesc>,    \
      paddle::framework::EmptyGradOpMaker<paddle::imperative::OpBase>);