cudnn_lstm_op.cu.cc 18.9 KB
Newer Older
P
phlrain 已提交
1
/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
L
liuhongyu 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 "paddle/fluid/framework/generator.h"
C
chengduozh 已提交
16
#include "paddle/fluid/framework/op_registry.h"
17
#include "paddle/fluid/operators/cudnn_lstm_cache.h"
C
chengduozh 已提交
18
#include "paddle/fluid/operators/math/math_function.h"
19
#include "paddle/fluid/operators/utils.h"
W
wanghuancoder 已提交
20 21 22 23 24 25 26

namespace paddle {
namespace platform {
class CUDADeviceContext;
struct CUDAPlace;
}  // namespace platform
}  // namespace paddle
L
liuhongyu 已提交
27 28 29 30 31 32 33

namespace paddle {
namespace operators {

using LoDTensor = framework::LoDTensor;
using Tensor = framework::Tensor;

G
GaoWei8 已提交
34 35 36 37 38 39 40 41 42 43 44 45 46 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 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93
template <typename T, typename Type>
bool is_continuous(const Type &weight_list) {
  bool continuous = true;
  for (size_t i = 0; i < weight_list.size() - 1; ++i) {
    auto *in_data = weight_list[i]->template data<T>();
    auto *in_after_data = weight_list[i + 1]->template data<T>();
    auto in_size = weight_list[i]->numel();
    bool temp = in_data + in_size == in_after_data;
    continuous = continuous && temp;
  }
  return continuous;
}

int size_sum(const std::vector<const Tensor *> &weight_list) {
  int size = 0;
  for (size_t i = 0; i < weight_list.size(); ++i) {
    auto in_size = weight_list[i]->numel();
    size += in_size;
  }
  return size;
}

template <typename T>
void weight_to_tensor(const platform::Place &place, cudaStream_t stream,
                      const std::vector<const Tensor *> &weight_list,
                      Tensor *weight) {
  auto weight_data = weight->data<T>();
  int weight_offset = 0;
  for (size_t i = 0; i < weight_list.size(); ++i) {
    const T *in_data = weight_list[i]->data<T>();
    auto in_size = weight_list[i]->numel();

    memory::Copy(BOOST_GET_CONST(platform::CUDAPlace, weight->place()),
                 weight_data + weight_offset,
                 BOOST_GET_CONST(platform::CUDAPlace, weight_list[i]->place()),
                 in_data, in_size * sizeof(T), stream);
    weight_offset += in_size;
  }
}

template <typename T>
void weight_to_tensor_list(const platform::Place &place, cudaStream_t stream,
                           std::vector<Tensor *> *weight_grad,
                           const std::vector<const Tensor *> &weight_input,
                           const Tensor *weight) {
  int weight_offset = 0;
  auto *weight_data = weight->data<T>();
  for (size_t i = 0; i < weight_input.size(); ++i) {
    auto in_size = weight_input[i]->numel();
    T *weight_grad_data = (*weight_grad)[i]->mutable_data<T>(place);
    const T *src = weight_data + weight_offset;

    memory::Copy(
        BOOST_GET_CONST(platform::CUDAPlace, (*weight_grad)[i]->place()),
        weight_grad_data, BOOST_GET_CONST(platform::CUDAPlace, weight->place()),
        src, in_size * sizeof(T), stream);
    weight_offset += in_size;
  }
}

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
template <typename T>
void LSTMInferece(const bool &has_seq_length, const cudnnHandle_t &handle,
                  const int &seq_length, ScopedRNNBase *rnn, const T *x_data,
                  const T *init_h_data, const T *init_c_data, const T *w_data,
                  T *out_data, T *last_h_data, T *last_c_data,
                  framework::Tensor *workspace_data,
                  const size_t &workspace_size) {
  if (!has_seq_length) {
    // for inference
    // This interface is used when the input/output is unpadded.
    PADDLE_ENFORCE_CUDA_SUCCESS(platform::dynload::cudnnRNNForwardInference(
        handle, rnn->rnn_desc(), seq_length, rnn->x_descs(), x_data,
        rnn->init_h_desc(), init_h_data, rnn->init_c_desc(), init_c_data,
        rnn->weight_desc(), w_data, rnn->y_descs(), out_data,
        rnn->last_h_desc(), last_h_data, rnn->last_c_desc(), last_c_data,
        workspace_data->data<uint8_t>(), workspace_size));
  } else {
#if CUDNN_VERSION >= 7201
    // for inference
    // This interface is used when the input/output is padded.
    PADDLE_ENFORCE_CUDA_SUCCESS(platform::dynload::cudnnRNNForwardInferenceEx(
        handle, rnn->rnn_desc(), rnn->x_seq_desc(), x_data, rnn->init_h_desc(),
        init_h_data, rnn->init_c_desc(), init_c_data, rnn->weight_desc(),
        w_data, rnn->y_seq_desc(), out_data, rnn->last_h_desc(), last_h_data,
        rnn->last_c_desc(), last_c_data, nullptr, nullptr, nullptr, nullptr,
        nullptr, nullptr, nullptr, nullptr, workspace_data->data<uint8_t>(),
        workspace_size));
#else
    // CUDNN VERSION has to >=7.2.1
    PADDLE_THROW(platform::errors::Unavailable(
        "The padded input is supported by "
        "cudnnRNNForwardInferenceEx, but it only works when "
        "the version of cudnn is larger than 7.2.1"));
#endif
  }
}

C
chengduozh 已提交
131
template <typename T>
L
liuhongyu 已提交
132 133 134 135 136 137 138 139
class CudnnLSTMGPUKernel : public framework::OpKernel<T> {
 public:
  void Compute(const framework::ExecutionContext &ctx) const override {
    const Tensor *x = ctx.Input<Tensor>("Input");
    const Tensor *init_h = ctx.Input<Tensor>("InitH");
    const Tensor *init_c = ctx.Input<Tensor>("InitC");

    Tensor *out = ctx.Output<Tensor>("Out");
G
GaoWei8 已提交
140 141 142 143
    Tensor *last_h = ctx.Output<Tensor>("LastH");
    Tensor *last_c = ctx.Output<Tensor>("LastC");
    Tensor *reserve = ctx.Output<Tensor>("Reserve");
    Tensor *state_out = ctx.Output<Tensor>("StateOut");
L
liuhongyu 已提交
144 145 146 147 148 149 150 151 152 153 154 155 156 157

    const T *x_data = x->data<T>();
    const T *init_h_data = init_h->data<T>();
    const T *init_c_data = init_c->data<T>();

    T *out_data = out->mutable_data<T>(ctx.GetPlace());
    T *last_h_data = last_h->mutable_data<T>(ctx.GetPlace());
    T *last_c_data = last_c->mutable_data<T>(ctx.GetPlace());

    float dropout_prob = ctx.Attr<float>("dropout_prob");
    bool is_bidirec = ctx.Attr<bool>("is_bidirec");
    int hidden_size = ctx.Attr<int>("hidden_size");
    int num_layers = ctx.Attr<int>("num_layers");
    bool is_test = ctx.Attr<bool>("is_test");
G
GaoWei8 已提交
158
    int seed = ctx.Attr<int>("seed");
159

160 161 162 163 164 165 166 167 168 169 170 171 172 173 174
    if (!is_test) {
      int device_id =
          BOOST_GET_CONST(platform::CUDAPlace, ctx.GetPlace()).GetDeviceId();
      auto gen_cuda = framework::GetDefaultCUDAGenerator(device_id);
      if (gen_cuda->GetIsInitPy() && seed == 0) {
        // If perform `manual_seed` in python and inner seed is not specified
        // (equals 0), use global generator generated seed.
        seed = static_cast<int>(gen_cuda->Random64());
      } else if (seed == 0) {
        // use random generated seed
        std::random_device rd;
        seed = rd();
      }  // else use `ctx.Attr<int>("seed")` specified seed
    }

175 176 177 178 179 180
    bool has_seq_length = ctx.HasInput("SequenceLength");
    std::vector<int> SequenceLength;
    if (has_seq_length) {
      auto *sequence_length = ctx.Input<Tensor>("SequenceLength");
      SequenceLength = operators::GetDataFromTensor<int>(sequence_length);
    }
L
liuhongyu 已提交
181 182 183 184

    auto &dev_ctx = ctx.template device_context<platform::CUDADeviceContext>();
    auto handle = dev_ctx.cudnn_handle();

G
GaoWei8 已提交
185 186 187 188
    int seq_length = x->dims()[0];
    int batch_size = x->dims()[1];
    int input_size = x->dims()[2];
    bool state_initialized = state_out->IsInitialized() ? true : false;
G
GaoWei8 已提交
189

G
GaoWei8 已提交
190
    size_t workspace_size;
G
GaoWei8 已提交
191
    size_t reserve_size;
G
GaoWei8 已提交
192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212
    Tensor weight_whole;
    T *w_data = nullptr;
    int weight_numel;
    bool w_initialized = false;
    auto place = ctx.GetPlace();
    auto stream = reinterpret_cast<const platform::CUDADeviceContext &>(
                      ctx.device_context())
                      .stream();
    if (is_test && ctx.HasInput("W")) {
      auto *W = ctx.Input<Tensor>("W");
      w_initialized = W->IsInitialized() ? true : false;
      weight_numel = W->numel();
    }
    if (!w_initialized) {
      auto weight_list = ctx.MultiInput<framework::Tensor>("WeightList");
      bool continuous =
          is_continuous<T, std::vector<const Tensor *>>(weight_list);
      weight_numel = size_sum(weight_list);

      if (!continuous) {
        LOG_FIRST_N(WARNING, 2)
213 214 215
            << "If the memory space of the Input WeightList is not continuous, "
               "less efficient calculation will be called. Please call "
               "flatten_parameters() to make the input memory continuous.";
G
GaoWei8 已提交
216 217 218
        weight_whole.mutable_data<T>({weight_numel}, place);
        weight_to_tensor<T>(place, stream, weight_list, &weight_whole);
        w_data = weight_whole.data<T>();
219 220 221 222 223 224 225 226 227 228 229 230 231
        if (is_test) {  // maybe also reset small weights' ptr for training
          int offset = 0;
          for (size_t i = 0; i < weight_list.size(); ++i) {
            size_t len = weight_list[i]->numel();
            auto dim = weight_list[i]->dims();
            const_cast<Tensor *>(weight_list[i])
                ->ShareDataWith(
                    weight_whole.Slice(static_cast<int64_t>(offset),
                                       static_cast<int64_t>(offset + len)))
                .Resize(dim);
            offset += len;
          }
        }
G
GaoWei8 已提交
232 233 234 235 236 237 238
      } else {
        w_data = const_cast<T *>(weight_list[0]->data<T>());
      }
    } else {
      auto *W = ctx.Input<Tensor>("W");
      w_data = const_cast<T *>(W->data<T>());
    }
G
GaoWei8 已提交
239

240 241 242 243
    ScopedRNNBase rnn(seq_length, batch_size, input_size, hidden_size,
                      num_layers, dropout_prob, seed, weight_numel,
                      state_initialized, is_bidirec);
    rnn.Create<T>(handle, ctx.GetPlace(), SequenceLength, &workspace_size,
G
GaoWei8 已提交
244 245 246
                  &reserve_size, state_out);

    framework::Tensor workspace_data_;
247 248
    workspace_data_.mutable_data<uint8_t>(
        {static_cast<int64_t>(workspace_size)}, ctx.GetPlace());
G
GaoWei8 已提交
249 250 251

    auto *reserve_data = reserve->mutable_data<uint8_t>(
        {static_cast<int64_t>(reserve_size)}, ctx.GetPlace());
L
liuhongyu 已提交
252 253

    if (is_test) {
254 255 256
      LSTMInferece<T>(has_seq_length, handle, seq_length, &rnn, x_data,
                      init_h_data, init_c_data, w_data, out_data, last_h_data,
                      last_c_data, &workspace_data_, workspace_size);
L
liuhongyu 已提交
257
    } else {
258
      if (!has_seq_length) {
G
GaoWei8 已提交
259 260 261
        // for train
        // This interface is used when the input/output is unpadded.
        PADDLE_ENFORCE_CUDA_SUCCESS(platform::dynload::cudnnRNNForwardTraining(
262 263 264 265
            handle, rnn.rnn_desc(), seq_length, rnn.x_descs(), x_data,
            rnn.init_h_desc(), init_h_data, rnn.init_c_desc(), init_c_data,
            rnn.weight_desc(), w_data, rnn.y_descs(), out_data,
            rnn.last_h_desc(), last_h_data, rnn.last_c_desc(), last_c_data,
G
GaoWei8 已提交
266 267 268 269 270 271 272 273
            workspace_data_.data<uint8_t>(), workspace_size, reserve_data,
            reserve_size));
      } else {
#if CUDNN_VERSION >= 7201
        // for train
        // This interface is used when the input/output is padded.
        PADDLE_ENFORCE_CUDA_SUCCESS(
            platform::dynload::cudnnRNNForwardTrainingEx(
274 275 276 277 278 279 280
                handle, rnn.rnn_desc(), rnn.x_seq_desc(), x_data,
                rnn.init_h_desc(), init_h_data, rnn.init_c_desc(), init_c_data,
                rnn.weight_desc(), w_data, rnn.y_seq_desc(), out_data,
                rnn.last_h_desc(), last_h_data, rnn.last_c_desc(), last_c_data,
                nullptr, nullptr, nullptr, nullptr, nullptr, nullptr, nullptr,
                nullptr, workspace_data_.data<uint8_t>(), workspace_size,
                reserve_data, reserve_size));
G
GaoWei8 已提交
281
#else
282 283 284 285
        PADDLE_THROW(platform::errors::Unavailable(
            "The padded input is supported by "
            "cudnnRNNForwardTrainingEx, but it only works when "
            "the version of cudnn is larger than 7.2.1"));
G
GaoWei8 已提交
286 287
#endif
      }
L
liuhongyu 已提交
288 289 290 291
    }
  }
};

C
chengduozh 已提交
292
template <typename T>
L
liuhongyu 已提交
293 294 295 296 297 298
class CudnnLSTMGPUGradKernel : public framework::OpKernel<T> {
 public:
  void Compute(const framework::ExecutionContext &ctx) const override {
    auto *input = ctx.Input<Tensor>("Input");
    auto *init_h = ctx.Input<Tensor>("InitH");
    auto *init_c = ctx.Input<Tensor>("InitC");
G
GaoWei8 已提交
299 300
    auto *reserve = ctx.Input<Tensor>("Reserve");
    auto *state_out = ctx.Input<Tensor>("StateOut");
G
GaoWei8 已提交
301
    auto weight_list = ctx.MultiInput<Tensor>("WeightList");
G
GaoWei8 已提交
302

L
liuhongyu 已提交
303 304
    auto *out = ctx.Input<Tensor>("Out");
    auto *out_grad = ctx.Input<Tensor>(framework::GradVarName("Out"));
G
GaoWei8 已提交
305 306
    auto *last_h_grad = ctx.Input<Tensor>(framework::GradVarName("LastH"));
    auto *last_c_grad = ctx.Input<Tensor>(framework::GradVarName("LastC"));
L
liuhongyu 已提交
307 308 309 310

    auto *in_grad = ctx.Output<Tensor>(framework::GradVarName("Input"));
    auto *init_h_grad = ctx.Output<Tensor>(framework::GradVarName("InitH"));
    auto *init_c_grad = ctx.Output<Tensor>(framework::GradVarName("InitC"));
G
GaoWei8 已提交
311 312
    auto weight_grad_list = ctx.MultiOutput<framework::Tensor>(
        framework::GradVarName("WeightList"));
L
liuhongyu 已提交
313 314 315 316 317 318 319 320

    auto &dev_ctx = ctx.template device_context<platform::CUDADeviceContext>();
    auto handle = dev_ctx.cudnn_handle();

    auto input_dims = input->dims();
    auto init_h_dims = init_h->dims();
    auto init_c_dims = init_c->dims();

G
GaoWei8 已提交
321 322 323 324 325 326
    auto *init_h_data = init_h->data<T>();
    auto *init_c_data = init_c->data<T>();
    auto *out_data = out->data<T>();
    auto *out_grad_data = out_grad->data<T>();
    auto *last_h_grad_data = last_h_grad->data<T>();
    auto *last_c_grad_data = last_c_grad->data<T>();
L
liuhongyu 已提交
327

G
GaoWei8 已提交
328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347
    auto place = ctx.GetPlace();
    int weight_numel = size_sum(weight_list);
    bool continuous =
        is_continuous<T, std::vector<const Tensor *>>(weight_list);

    auto stream = reinterpret_cast<const platform::CUDADeviceContext &>(
                      ctx.device_context())
                      .stream();
    Tensor weight_whole;
    T *weight_data = nullptr;

    if (!continuous) {
      weight_whole.mutable_data<T>({weight_numel}, place);
      weight_to_tensor<T>(place, stream, weight_list, &weight_whole);
      weight_data = weight_whole.data<T>();
    } else {
      weight_data = const_cast<T *>(weight_list[0]->data<T>());
    }

    Tensor weight_grad;
G
GaoWei8 已提交
348
    math::SetConstant<paddle::platform::CUDADeviceContext, T> zero;
G
GaoWei8 已提交
349 350 351 352 353 354 355 356 357 358 359 360 361 362
    weight_grad.mutable_data<T>({weight_numel}, ctx.GetPlace());
    zero(dev_ctx, &weight_grad, static_cast<T>(0.0));
    T *weight_grad_data = weight_grad.data<T>();

    int offset = 0;
    for (size_t i = 0; i < weight_grad_list.size(); ++i) {
      size_t len = weight_grad_list[i]->numel();
      auto dim = weight_grad_list[i]->dims();
      weight_grad_list[i]
          ->ShareDataWith(weight_grad.Slice(static_cast<int64_t>(offset),
                                            static_cast<int64_t>(offset + len)))
          .Resize(dim);
      offset += len;
    }
L
liuhongyu 已提交
363

G
GaoWei8 已提交
364 365
    in_grad->mutable_data<T>(input_dims, ctx.GetPlace());
    auto *in_grad_data = in_grad->data<T>();
L
liuhongyu 已提交
366

G
GaoWei8 已提交
367 368
    if (init_h_grad) init_h_grad->mutable_data<T>(init_h_dims, ctx.GetPlace());
    auto *init_h_grad_data = init_h_grad ? init_h_grad->data<T>() : nullptr;
L
liuhongyu 已提交
369

G
GaoWei8 已提交
370 371
    if (init_c_grad) init_c_grad->mutable_data<T>(init_c_dims, ctx.GetPlace());
    auto *init_c_grad_data = init_c_grad ? init_c_grad->data<T>() : nullptr;
L
liuhongyu 已提交
372

G
GaoWei8 已提交
373 374 375 376 377
    float dropout_prob = ctx.Attr<float>("dropout_prob");
    bool is_bidirec = ctx.Attr<bool>("is_bidirec");
    int hidden_size = ctx.Attr<int>("hidden_size");
    int num_layers = ctx.Attr<int>("num_layers");
    int seed = ctx.Attr<int>("seed");
378 379 380 381 382 383 384

    bool has_seq_length = ctx.HasInput("SequenceLength");
    std::vector<int> SequenceLength;
    if (has_seq_length) {
      auto *sequence_length = ctx.Input<Tensor>("SequenceLength");
      SequenceLength = operators::GetDataFromTensor<int>(sequence_length);
    }
G
GaoWei8 已提交
385

G
GaoWei8 已提交
386 387 388
    int seq_length = input_dims[0];
    int batch_size = input->dims()[1];
    int input_size = input->dims()[2];
G
GaoWei8 已提交
389

G
GaoWei8 已提交
390
    size_t workspace_size;
G
GaoWei8 已提交
391
    size_t reserve_size;
G
GaoWei8 已提交
392

393 394 395
    ScopedRNNBase rnn(seq_length, batch_size, input_size, hidden_size,
                      num_layers, dropout_prob, seed, weight_numel, true,
                      is_bidirec);
G
GaoWei8 已提交
396

397
    rnn.Create<T>(handle, ctx.GetPlace(), SequenceLength, &workspace_size,
G
GaoWei8 已提交
398 399 400
                  &reserve_size, const_cast<Tensor *>(state_out));

    framework::Tensor workspace_data_;
401 402
    workspace_data_.mutable_data<uint8_t>(
        {static_cast<int64_t>(workspace_size)}, ctx.GetPlace());
G
GaoWei8 已提交
403
    const uint8_t *reserve_data = reserve->data<uint8_t>();
L
liuhongyu 已提交
404

405
    if (!has_seq_length) {
G
GaoWei8 已提交
406 407
      // This interface is used when the input/output is unpadded.
      PADDLE_ENFORCE_CUDA_SUCCESS(platform::dynload::cudnnRNNBackwardData(
408 409 410 411 412 413 414
          handle, rnn.rnn_desc(), seq_length, rnn.y_descs(), out_data,
          rnn.y_descs(), out_grad_data, rnn.last_h_desc(), last_h_grad_data,
          rnn.last_c_desc(), last_c_grad_data, rnn.weight_desc(), weight_data,
          rnn.init_h_desc(), init_h_data, rnn.init_c_desc(), init_c_data,
          rnn.x_descs(), in_grad_data, rnn.init_h_desc(), init_h_grad_data,
          rnn.init_c_desc(), init_c_grad_data, workspace_data_.data<uint8_t>(),
          workspace_size, const_cast<uint8_t *>(reserve_data), reserve_size));
G
GaoWei8 已提交
415 416

      PADDLE_ENFORCE_CUDA_SUCCESS(platform::dynload::cudnnRNNBackwardWeights(
417 418 419
          handle, rnn.rnn_desc(), seq_length, rnn.x_descs(), input->data<T>(),
          rnn.init_h_desc(), init_h->data<T>(), rnn.y_descs(), out->data<T>(),
          workspace_data_.data<uint8_t>(), workspace_size, rnn.weight_desc(),
G
GaoWei8 已提交
420
          weight_grad_data, const_cast<uint8_t *>(reserve_data), reserve_size));
G
GaoWei8 已提交
421 422 423 424 425 426
    } else {
#if CUDNN_VERSION >= 7201
      // for train
      // This interface is used when the input/output is padded.
      PADDLE_ENFORCE_CUDA_SUCCESS(platform::dynload::cudnnRNNBackwardDataEx(
          handle, rnn.rnn_desc(), rnn.y_seq_desc(), out_data, rnn.y_seq_desc(),
427 428 429 430 431
          out_grad_data, nullptr, nullptr, rnn.last_h_desc(), last_h_grad_data,
          rnn.last_c_desc(), last_c_grad_data, rnn.weight_desc(), weight_data,
          rnn.init_h_desc(), init_h_data, rnn.init_c_desc(), init_c_data,
          rnn.x_seq_desc(), in_grad_data, rnn.init_h_desc(), init_h_grad_data,
          rnn.init_c_desc(), init_c_grad_data, nullptr, nullptr,
G
GaoWei8 已提交
432 433 434 435 436
          workspace_data_.data<uint8_t>(), workspace_size,
          const_cast<uint8_t *>(reserve_data), reserve_size));

      PADDLE_ENFORCE_CUDA_SUCCESS(platform::dynload::cudnnRNNBackwardWeightsEx(
          handle, rnn.rnn_desc(), rnn.x_seq_desc(), input->data<T>(),
437 438
          rnn.init_h_desc(), init_h->data<T>(), rnn.y_seq_desc(),
          out->data<T>(), workspace_data_.data<uint8_t>(), workspace_size,
G
GaoWei8 已提交
439
          rnn.weight_desc(), weight_grad_data,
440
          const_cast<uint8_t *>(reserve_data), reserve_size));
G
GaoWei8 已提交
441
#else
442 443 444 445
      PADDLE_THROW(platform::errors::Unavailable(
          "The padded input of rnn is supported by cudnnRNNBackwardDataEx, "
          "cudnnRNNBackwardWeightsEx, but it only works when the version "
          "of cudnn is larger than 7.2.1"));
G
GaoWei8 已提交
446 447
#endif
    }
L
liuhongyu 已提交
448 449 450 451 452 453 454
  }
};

}  // namespace operators
}  // namespace paddle

namespace ops = paddle::operators;
G
GaoWei8 已提交
455 456 457 458
REGISTER_OP_CUDA_KERNEL(cudnn_lstm, ops::CudnnLSTMGPUKernel<float>,
                        ops::CudnnLSTMGPUKernel<double>);
REGISTER_OP_CUDA_KERNEL(cudnn_lstm_grad, ops::CudnnLSTMGPUGradKernel<float>,
                        ops::CudnnLSTMGPUGradKernel<double>);