blob: 984e8d68c09189619157f7162338d667f800b1bf [file] [log] [blame]
Giorgio Arena657bdb32018-04-26 18:52:01 +01001/*
Michalis Spyrouebcebf12020-10-21 00:04:14 +01002 * Copyright (c) 2018-2020 Arm Limited.
Giorgio Arena657bdb32018-04-26 18:52:01 +01003 *
4 * SPDX-License-Identifier: MIT
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to
8 * deal in the Software without restriction, including without limitation the
9 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
10 * sell copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in all
14 * copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22 * SOFTWARE.
23 */
Michalis Spyrouf4643372019-11-29 16:17:13 +000024#ifndef ARM_COMPUTE_NECONVERTFULLYCONNECTEDWEIGHTS_H
25#define ARM_COMPUTE_NECONVERTFULLYCONNECTEDWEIGHTS_H
Giorgio Arena657bdb32018-04-26 18:52:01 +010026
Giorgio Arena657bdb32018-04-26 18:52:01 +010027#include "arm_compute/runtime/IFunction.h"
Michalis Spyrou1a569a32019-09-10 17:20:34 +010028#include "arm_compute/runtime/ITransformWeights.h"
Giorgio Arena657bdb32018-04-26 18:52:01 +010029#include "arm_compute/runtime/NEON/NEScheduler.h"
Michalis Spyrou1a569a32019-09-10 17:20:34 +010030#include "arm_compute/runtime/Tensor.h"
Michalis Spyrouebcebf12020-10-21 00:04:14 +010031#include <memory>
Giorgio Arena657bdb32018-04-26 18:52:01 +010032
33namespace arm_compute
34{
Georgios Pinitas33843562019-12-10 13:33:18 +000035// Forward declarations
Giorgio Arena657bdb32018-04-26 18:52:01 +010036class ITensor;
Michalis Spyrouebcebf12020-10-21 00:04:14 +010037class NEConvertFullyConnectedWeightsKernel;
Giorgio Arena657bdb32018-04-26 18:52:01 +010038
39/** Basic function to run @ref NEConvertFullyConnectedWeightsKernel. */
40class NEConvertFullyConnectedWeights : public IFunction
41{
42public:
43 /** Default constructor */
44 NEConvertFullyConnectedWeights();
Michalis Spyrouebcebf12020-10-21 00:04:14 +010045 /** Prevent instances of this class from being copied (As this class contains pointers) */
46 NEConvertFullyConnectedWeights(const NEConvertFullyConnectedWeights &) = delete;
47 /** Prevent instances of this class from being copied (As this class contains pointers) */
48 NEConvertFullyConnectedWeights &operator=(const NEConvertFullyConnectedWeights &) = delete;
49 /** Prevent instances of this class from being moved (As this class contains non movable objects) */
50 NEConvertFullyConnectedWeights(NEConvertFullyConnectedWeights &&) = delete;
51 /** Prevent instances of this class from being moved (As this class contains non movable objects) */
52 NEConvertFullyConnectedWeights &operator=(NEConvertFullyConnectedWeights &&) = delete;
53 /** Default destructor */
54 ~NEConvertFullyConnectedWeights();
Giorgio Arena657bdb32018-04-26 18:52:01 +010055 /** Initialize the function.
56 *
Georgios Pinitas33843562019-12-10 13:33:18 +000057 * @param[in] input Source weights tensor to convert. Must be 2 dimensional. Data types supported: All.
Giorgio Arena657bdb32018-04-26 18:52:01 +010058 * @param[out] output The converted weights tensor. Shape and Data Type: Same as @p input.
Georgios Pinitas7d66a8e2018-07-17 12:28:42 +010059 * @param[in] original_input_shape Shape of the original input tensor (the one entering fully connected layer).
Giorgio Arena657bdb32018-04-26 18:52:01 +010060 * @param[in] data_layout The data layout the weights have been trained in.
61 */
62 void configure(const ITensor *input, ITensor *output, const TensorShape &original_input_shape, DataLayout data_layout);
63 /** Static function to check if given info will lead to a valid configuration of @ref NEConvertFullyConnectedWeights
64 *
Georgios Pinitas33843562019-12-10 13:33:18 +000065 * @param[in] input Source weights tensor info to convert. Must be 2 dimensional. Data types supported: All.
Giorgio Arena657bdb32018-04-26 18:52:01 +010066 * @param[in] output The converted weights tensor info. Shape and Data Type: Same as @p input.
Georgios Pinitas7d66a8e2018-07-17 12:28:42 +010067 * @param[in] original_input_shape Shape of the original input tensor (the one entering fully connected layer).
Giorgio Arena657bdb32018-04-26 18:52:01 +010068 * @param[in] data_layout The data layout the weights have been trained in.
Michalis Spyrou1a569a32019-09-10 17:20:34 +010069 *
70 * @return A Status
Giorgio Arena657bdb32018-04-26 18:52:01 +010071 */
72 static Status validate(const ITensorInfo *input, const ITensorInfo *output, const TensorShape &original_input_shape, DataLayout data_layout);
73
74 // Inherited methods overriden:
75 void run() override;
76
77private:
Michalis Spyrouebcebf12020-10-21 00:04:14 +010078 std::unique_ptr<NEConvertFullyConnectedWeightsKernel> _kernel;
Giorgio Arena657bdb32018-04-26 18:52:01 +010079};
Michalis Spyrou1a569a32019-09-10 17:20:34 +010080
81namespace weights_transformations
82{
83/** Basic function to run @ref NEConvertFullyConnectedWeightsKernel. */
84class NEConvertFullyConnectedWeightsManaged : public ITransformWeights
85{
86public:
87 void run() override
88 {
89 _output.allocator()->allocate();
90 _func.run();
91 _reshape_run = true;
92 }
93
94 void release() override
95 {
96 _output.allocator()->free();
97 }
98
99 ITensor *get_weights() override
100 {
101 return &_output;
102 }
103
104 uint32_t uid() override
105 {
106 return _uid;
107 }
108
109 void configure(const ITensor *input, const TensorShape &original_input_shape, DataLayout data_layout)
110 {
111 _func.configure(input, &_output, original_input_shape, data_layout);
112 }
113
114private:
115 static constexpr uint32_t _uid = 0x4;
116 Tensor _output{};
117 NEConvertFullyConnectedWeights _func{};
118};
119} // namespace weights_transformations
120} // namespace arm_compute
Michalis Spyrouf4643372019-11-29 16:17:13 +0000121#endif /* ARM_COMPUTE_NECONVERTFULLYCONNECTEDWEIGHTS_H */