1 /* Copyright 2015 The TensorFlow Authors. All Rights Reserved. 2 3 Licensed under the Apache License, Version 2.0 (the "License"); 4 you may not use this file except in compliance with the License. 5 You may obtain a copy of the License at 6 7 http://www.apache.org/licenses/LICENSE-2.0 8 9 Unless required by applicable law or agreed to in writing, software 10 distributed under the License is distributed on an "AS IS" BASIS, 11 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 12 See the License for the specific language governing permissions and 13 limitations under the License. 14 ==============================================================================*/ 15 16 #ifndef TENSORFLOW_CORE_KERNELS_BIAS_OP_H_ 17 #define TENSORFLOW_CORE_KERNELS_BIAS_OP_H_ 18 // Functor definition for BiasOp, must be compilable by nvcc. 19 20 #include "third_party/eigen3/unsupported/Eigen/CXX11/Tensor" 21 #include "tensorflow/core/framework/tensor_types.h" 22 23 namespace tensorflow { 24 namespace functor { 25 26 // Functor used by BiasOp to do the computations. 27 template <typename Device, typename T> 28 struct Bias { 29 // Add "bias" to "input", repeating "bias". operatorBias30 void operator()(const Device& d, typename TTypes<T>::ConstFlat input, 31 typename TTypes<T>::ConstVec bias, 32 typename TTypes<T>::Flat output) { 33 const Eigen::Index rest_size = input.size() / bias.dimension(0); 34 Eigen::DSizes<Eigen::Index, 1> bcast(rest_size); 35 MaybeWith32BitIndexing<Device>( 36 [&](auto input32, auto bias32, auto output32, const auto& bcast32) { 37 output32.device(d) = input32 + bias32.broadcast(bcast32); 38 }, 39 input, bias, output, bcast); 40 } 41 42 // NCHW layout, repeating on the first dimension, broadcasting on the last 43 // dimension. operatorBias44 void operator()(const Device& d, typename TTypes<T>::ConstMatrix input, 45 typename TTypes<T>::ConstMatrix bias1, // shape [C, 1]. 46 typename TTypes<T>::Matrix output) { 47 const Eigen::Index rest_size = input.dimension(0) / bias1.dimension(0); 48 Eigen::DSizes<Eigen::Index, 2> bcast(rest_size, input.dimension(1)); 49 MaybeWith32BitIndexing<Device>( 50 [&](auto input32, auto bias32, auto output32, const auto& bcast32) { 51 output32.device(d) = input32 + bias32.broadcast(bcast32); 52 }, 53 input, bias1, output, bcast); 54 } 55 }; 56 57 } // namespace functor 58 } // namespace tensorflow 59 60 #endif // TENSORFLOW_CORE_KERNELS_BIAS_OP_H_ 61