1 /*
2 * Copyright (c) 2018-2021 Arm Limited.
3 *
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 */
24
25 #include "arm_compute/runtime/NEON/functions/NERNNLayer.h"
26
27 #include "arm_compute/core/Error.h"
28 #include "arm_compute/core/TensorInfo.h"
29 #include "arm_compute/core/Types.h"
30 #include "arm_compute/core/Validate.h"
31 #include "arm_compute/core/utils/misc/ShapeCalculator.h"
32 #include "arm_compute/runtime/NEON/NEScheduler.h"
33 #include "src/common/utils/Log.h"
34
35 namespace arm_compute
36 {
37 NERNNLayer::~NERNNLayer() = default;
38
NERNNLayer(std::shared_ptr<IMemoryManager> memory_manager)39 NERNNLayer::NERNNLayer(std::shared_ptr<IMemoryManager> memory_manager)
40 : _memory_group(std::move(memory_manager)), _gemm_state_f(), _add_f(), _activation(), _fully_connected(memory_manager), _copy_f(), _fully_connected_out(), _gemm_output(), _add_output(),
41 _is_prepared(false)
42 {
43 }
44
validate(const ITensorInfo * input,const ITensorInfo * weights,const ITensorInfo * recurrent_weights,const ITensorInfo * bias,const ITensorInfo * hidden_state,const ITensorInfo * output,const ActivationLayerInfo & info)45 Status NERNNLayer::validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *recurrent_weights, const ITensorInfo *bias, const ITensorInfo *hidden_state,
46 const ITensorInfo *output, const ActivationLayerInfo &info)
47 {
48 ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, weights, recurrent_weights, bias, hidden_state, output);
49 ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_NOT_IN(input, DataType::F16, DataType::F32);
50
51 const int idx_width = get_data_layout_dimension_index(input->data_layout(), DataLayoutDimension::WIDTH);
52 const int idx_height = get_data_layout_dimension_index(input->data_layout(), DataLayoutDimension::HEIGHT);
53 ARM_COMPUTE_RETURN_ERROR_ON(input->dimension(idx_width) != weights->dimension(idx_width));
54 ARM_COMPUTE_RETURN_ERROR_ON(input->num_dimensions() != 2);
55 ARM_COMPUTE_RETURN_ERROR_ON(weights->dimension(idx_height) != recurrent_weights->dimension(idx_width));
56 ARM_COMPUTE_RETURN_ERROR_ON(recurrent_weights->dimension(idx_width) != recurrent_weights->dimension(idx_height));
57 ARM_COMPUTE_RETURN_ERROR_ON(bias->num_dimensions() != 1);
58 ARM_COMPUTE_RETURN_ERROR_ON(bias->dimension(idx_width) != weights->dimension(idx_height));
59 ARM_COMPUTE_RETURN_ERROR_ON(hidden_state->dimension(idx_width) != weights->dimension(idx_height));
60 ARM_COMPUTE_RETURN_ERROR_ON(hidden_state->dimension(idx_height) != input->dimension(idx_height));
61 ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DIMENSIONS(output->tensor_shape(), hidden_state->tensor_shape());
62
63 auto shape_info = TensorInfo(misc::shape_calculator::compute_rnn_shape(recurrent_weights, hidden_state->dimension(idx_height)), 1, input->data_type());
64
65 ARM_COMPUTE_RETURN_ON_ERROR(NEFullyConnectedLayer::validate(input, weights, bias, &shape_info));
66 ARM_COMPUTE_RETURN_ON_ERROR(NEArithmeticAddition::validate(&shape_info, &shape_info, &shape_info, ConvertPolicy::SATURATE));
67 ARM_COMPUTE_RETURN_ON_ERROR(NEActivationLayer::validate(&shape_info, &shape_info, info));
68
69 return Status{};
70 }
71
configure(const ITensor * input,const ITensor * weights,const ITensor * recurrent_weights,const ITensor * bias,ITensor * hidden_state,ITensor * output,ActivationLayerInfo & info)72 void NERNNLayer::configure(const ITensor *input, const ITensor *weights, const ITensor *recurrent_weights, const ITensor *bias, ITensor *hidden_state, ITensor *output,
73 ActivationLayerInfo &info)
74 {
75 ARM_COMPUTE_ERROR_ON_NULLPTR(input, weights, recurrent_weights, bias, hidden_state, output);
76 ARM_COMPUTE_ERROR_THROW_ON(NERNNLayer::validate(input->info(), weights->info(), recurrent_weights->info(), bias->info(), hidden_state->info(), output->info(), info));
77 ARM_COMPUTE_LOG_PARAMS(input, weights, recurrent_weights, bias, hidden_state, output, info);
78
79 const int idx_height = get_data_layout_dimension_index(input->info()->data_layout(), DataLayoutDimension::HEIGHT);
80 TensorShape shape = misc::shape_calculator::compute_rnn_shape(recurrent_weights->info(), hidden_state->info()->dimension(idx_height));
81
82 _is_prepared = false;
83
84 // Manage intermediate buffers and configure
85 _fully_connected_out.allocator()->init(TensorInfo(shape, 1, input->info()->data_type()));
86 _gemm_output.allocator()->init(TensorInfo(shape, 1, input->info()->data_type()));
87
88 // Manage intermediate buffers and configure
89 _memory_group.manage(&_fully_connected_out);
90 _fully_connected.configure(input, weights, bias, &_fully_connected_out);
91
92 _memory_group.manage(&_gemm_output);
93 _gemm_state_f.configure(hidden_state, recurrent_weights, nullptr, &_gemm_output, 1.f, 0.f);
94
95 _add_output.allocator()->init(TensorInfo(shape, 1, input->info()->data_type()));
96 _memory_group.manage(&_add_output);
97
98 _add_f.configure(&_fully_connected_out, &_gemm_output, &_add_output, ConvertPolicy::SATURATE);
99
100 _fully_connected_out.allocator()->allocate();
101 _gemm_output.allocator()->allocate();
102
103 _activation.configure(&_add_output, hidden_state, info);
104 _add_output.allocator()->allocate();
105
106 _copy_f.configure(hidden_state, output);
107 }
108
run()109 void NERNNLayer::run()
110 {
111 prepare();
112
113 MemoryGroupResourceScope scope_mg(_memory_group);
114
115 _fully_connected.run();
116
117 _gemm_state_f.run();
118
119 _add_f.run();
120 _activation.run();
121
122 // copy hidden out to output
123 _copy_f.run();
124 }
125
prepare()126 void NERNNLayer::prepare()
127 {
128 if(!_is_prepared)
129 {
130 _fully_connected.prepare();
131 _gemm_state_f.prepare();
132
133 _is_prepared = true;
134 }
135 }
136 } // namespace arm_compute
137