sgd_kernel.cc 7.6 KB
Newer Older
P
phlrain 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

#include "paddle/phi/kernels/sgd_kernel.h"
#include "paddle/fluid/operators/jit/kernels.h"
P
update  
phlrain 已提交
17 18
#include "paddle/phi/backends/cpu/cpu_context.h"
#include "paddle/phi/core/kernel_registry.h"
P
phlrain 已提交
19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 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 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116
#include "paddle/phi/kernels/funcs/eigen/common.h"

namespace phi {

template <typename T>
void sgd_dense_param_dense_grad_impl(const DenseTensor& param,
                                     const DenseTensor& learning_rate,
                                     const DenseTensor& grad,
                                     DenseTensor* param_out) {
  const auto sz = param_out->numel();
  paddle::operators::jit::sgd_attr_t attr(1, sz, 1, sz, 1);
  const T* lr = learning_rate.data<T>();
  const T* param_data = param.data<T>();
  const T* grad_data = grad.data<T>();
  int64_t rows_idx = 0;
  T* out_data = param_out->data<T>();

  auto sgd =
      paddle::operators::jit::KernelFuncs<paddle::operators::jit::SgdTuple<T>,
                                          phi::CPUPlace>::Cache()
          .At(attr);
  sgd(lr, param_data, grad_data, &rows_idx, out_data, &attr);
}

template <>
void sgd_dense_param_dense_grad_impl<phi::dtype::bfloat16>(
    const DenseTensor& param,
    const DenseTensor& learning_rate,
    const DenseTensor& grad,
    DenseTensor* param_out) {
  auto p = EigenVector<phi::dtype::bfloat16>::Flatten(param);
  auto g = EigenVector<phi::dtype::bfloat16>::Flatten(grad);
  auto o = EigenVector<phi::dtype::bfloat16>::Flatten(*param_out);
  const auto* lr = learning_rate.data<phi::dtype::bfloat16>();

  o = p - lr[0] * g;
}

template <typename T>
void sgd_dense_param_sparse_grad_impl(const DenseTensor& param,
                                      const DenseTensor& learning_rate,
                                      const SelectedRows& grad,
                                      DenseTensor* param_out) {
  const auto& grad_value = grad.value();
  const auto& grad_rows = grad.rows();
  const T* param_data = param.data<T>();
  const T* grad_data = grad_value.data<T>();
  const T* lr = learning_rate.data<T>();
  const int64_t* rows_data = grad_rows.data();
  T* out_data = param_out->data<T>();

  paddle::operators::jit::sgd_attr_t attr;
  attr.param_height = param_out->dims()[0];
  attr.param_width = param_out->numel() / attr.param_height;
  attr.grad_height = grad_rows.size();  // note: it is not grad->height()
  attr.grad_width = grad_value.numel() / attr.grad_height;
  attr.selected_rows_size = grad_rows.size();

  auto sgd =
      paddle::operators::jit::KernelFuncs<paddle::operators::jit::SgdTuple<T>,
                                          phi::CPUPlace>::Cache()
          .At(attr);
  sgd(lr, param_data, grad_data, rows_data, out_data, &attr);
}

template <>
void sgd_dense_param_sparse_grad_impl<phi::dtype::bfloat16>(
    const DenseTensor& param,
    const DenseTensor& learning_rate,
    const SelectedRows& grad,
    DenseTensor* param_out) {
  const auto& grad_value = grad.value();
  const auto& grad_rows = grad.rows();
  const auto grad_height = grad.height();
  const int64_t grad_val_height = static_cast<int64_t>(grad_rows.size());
  const auto grad_width = grad_value.numel() / grad_val_height;

  const auto* grad_data = grad_value.data<phi::dtype::bfloat16>();
  auto* out_data = param_out->data<phi::dtype::bfloat16>();
  const auto* lr = learning_rate.data<phi::dtype::bfloat16>();

  for (size_t i = 0; i < grad_rows.size(); ++i) {
    PADDLE_ENFORCE_LT(
        grad_rows[i],
        grad_height,
        phi::errors::OutOfRange(
            "Grad rows index value should be less than grad height."
            "Got [%s], but expected less than [%s]",
            grad_rows[i],
            grad_height));
    const int64_t row = grad_rows[i];
    for (int64_t j = 0; j < grad_width; ++j) {
      out_data[row * grad_width + j] -= lr[0] * grad_data[i * grad_width + j];
    }
  }
}

template <typename T, typename Context>
P
update  
phlrain 已提交
117 118 119 120 121 122 123 124
void SGDDenseKernel(const Context& dev_ctx,
                    const DenseTensor& param,
                    const DenseTensor& learning_rate,
                    const DenseTensor& grad,
                    paddle::optional<const DenseTensor&> master_param,
                    bool multi_precision,
                    DenseTensor* param_out,
                    DenseTensor* master_param_out) {
P
phlrain 已提交
125 126 127 128 129
  dev_ctx.template Alloc<T>(param_out);
  sgd_dense_param_dense_grad_impl<T>(param, learning_rate, grad, param_out);
}

template <typename T, typename Context>
P
update  
phlrain 已提交
130 131 132 133 134 135 136 137 138
void SGDDenseParamSparseGradKernel(
    const Context& dev_ctx,
    const DenseTensor& param,
    const DenseTensor& learning_rate,
    const SelectedRows& grad,
    paddle::optional<const DenseTensor&> master_param,
    bool multi_precision,
    DenseTensor* param_out,
    DenseTensor* master_param_out) {
P
phlrain 已提交
139 140 141 142 143
  dev_ctx.template Alloc<T>(param_out);
  sgd_dense_param_sparse_grad_impl<T>(param, learning_rate, grad, param_out);
}

template <typename T, typename Context>
P
update  
phlrain 已提交
144 145 146 147 148 149 150 151 152
void SGDSparseParamSparseGradKernel(
    const Context& dev_ctx,
    const SelectedRows& param,
    const DenseTensor& learning_rate,
    const SelectedRows& grad,
    paddle::optional<const SelectedRows&> master_param,
    bool multi_precision,
    SelectedRows* param_out,
    SelectedRows* master_param_out) {
P
phlrain 已提交
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
  // for distributed training, a sparse var may be empty,
  // just skip updating.
  if (grad.rows().size() == 0) {
    return;
  }

  auto param_row_width = param.value().dims()[1];
  auto grad_row_width = grad.value().dims()[1];
  PADDLE_ENFORCE_EQ(
      param_row_width,
      grad_row_width,
      phi::errors::InvalidArgument(
          "The param_row in SgdOP should have the same size with grad_row. "
          "But received param_row's width is [%s], and grad_row's width is "
          "[%s]",
          param_row_width,
          grad_row_width));

  const auto* lr = learning_rate.data<T>();
  const auto* grad_data = grad.value().data<T>();
  auto* out_data = param_out->mutable_value()->data<T>();
  for (size_t i = 0; i < grad.rows().size(); i++) {
    int64_t id_index = param_out->AutoGrownIndex(grad.rows()[i], false);
    PADDLE_ENFORCE_GE(
        id_index,
        static_cast<int64_t>(0),
        phi::errors::InvalidArgument(
            "The id in SgdOp should be >= 0. But recevied id_index is [%s]",
            id_index));
    for (int64_t j = 0; j < grad_row_width; j++) {
      out_data[id_index * grad_row_width + j] -=
          lr[0] * grad_data[i * grad_row_width + j];
    }
  }
}

}  // namespace phi
P
update  
phlrain 已提交
190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213

PD_REGISTER_KERNEL(sgd,
                   CPU,
                   ALL_LAYOUT,
                   phi::SGDDenseKernel,
                   phi::dtype::bfloat16,
                   float,
                   double) {}

PD_REGISTER_KERNEL(sgd_dense_param_sparse_grad,
                   CPU,
                   ALL_LAYOUT,
                   phi::SGDDenseParamSparseGradKernel,
                   phi::dtype::bfloat16,
                   float,
                   double) {}

PD_REGISTER_KERNEL(sgd_sparse_param_sparse_grad,
                   CPU,
                   ALL_LAYOUT,
                   phi::SGDSparseParamSparseGradKernel,
                   phi::dtype::bfloat16,
                   float,
                   double) {}