Compute Library
 21.02
GCDropoutLayerKernel.cpp
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2017-2020 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 
26 
27 #include "arm_compute/core/Error.h"
33 #include "arm_compute/core/Types.h"
37 #include "support/StringSupport.h"
38 
39 #include <cmath>
40 #include <random>
41 #include <tuple>
42 
43 using namespace arm_compute;
44 
46  : _input(nullptr), _mask(nullptr), _output(nullptr), _num_elems_processed_per_iteration(0)
47 {
48 }
49 
50 void GCDropoutLayerKernel::configure(const IGCTensor *input, IGCTensor *mask, IGCTensor *output, float ratio, bool forward)
51 {
54 
55  _input = input;
56  _mask = mask;
57  _output = output;
58 
59  std::set<std::string> build_opts;
60  std::string dt_name = (input->info()->data_type() == DataType::F32) ? "DATA_TYPE_FP32" : "DATA_TYPE_FP16";
61  std::string fporbp = forward ? "FORWARD" : "BACKWARD";
62  std::random_device rd;
63  std::mt19937 mt(rd());
64  std::uniform_real_distribution<float> dist(0.f, 1.f);
65 
66  build_opts.emplace("#define LOCAL_SIZE_X " + support::cpp11::to_string(1));
67  build_opts.emplace("#define LOCAL_SIZE_Y " + support::cpp11::to_string(1));
68  build_opts.emplace("#define LOCAL_SIZE_Z " + support::cpp11::to_string(1));
69  build_opts.emplace("#define RATIO " + support::cpp11::to_string(ratio));
70  build_opts.emplace("#define SCALE " + support::cpp11::to_string(1. / (1. - ratio)));
71  build_opts.emplace("#define SEED " + support::cpp11::to_string(dist(mt)));
72  build_opts.emplace("#define " + dt_name);
73  build_opts.emplace("#define " + fporbp);
74 
75  _num_elems_processed_per_iteration = 4 / input->info()->element_size();
76 
77  // Create kernel
78  _kernel = static_cast<GCKernel>(GCKernelLibrary::get().create_kernel("dropout", build_opts));
79 
80  // Configure kernel window
81  Window win = calculate_max_window(*input->info(), Steps(_num_elems_processed_per_iteration));
82 
83  output->info()->set_valid_region(ValidRegion(Coordinates(), output->info()->tensor_shape()));
84 
85  IGCKernel::configure(win);
86 }
87 
89 {
92 
93  _kernel.use();
94 
96 
97  do
98  {
99  unsigned int idx = 0;
100 
101  add_3D_tensor_argument(idx, _input, 1, slice);
102  add_3D_tensor_argument(idx, _mask, 2, slice);
103  add_3D_tensor_argument(idx, _output, 3, slice);
104 
105  _kernel.update_shader_params();
106  enqueue(*this, slice);
107  }
108  while(window.slide_window_slice_3D(slice));
109 }
Window calculate_max_window(const ValidRegion &valid_region, const Steps &steps, bool skip_border, BorderSize border_size)
const Window & window() const
The maximum window the kernel can be executed on.
Definition: IKernel.cpp:28
void add_3D_tensor_argument(unsigned int &idx, const IGCTensor *tensor, const unsigned int binding_point, const Window &window)
Add the passed 3D tensor&#39;s parameters to the object&#39;s kernel&#39;s arguments starting from the index idx...
Definition: IGCKernel.cpp:132
void enqueue(IGCKernel &kernel, const Window &window, const gles::NDRange &lws=gles::NDRange(1U, 1U, 1U))
Add the kernel to the command queue with the given window.
Definition: IGCKernel.cpp:41
std::string to_string(T &&value)
Convert integer and float values to string.
virtual DataType data_type() const =0
Data type used for each element of the tensor.
1 channel, 1 F32 per channel
Interface for GLES Compute tensor.
Definition: IGCTensor.h:35
GCKernel class.
GCDropoutLayerKernel()
Default constructor.
Copyright (c) 2017-2021 Arm Limited.
virtual void set_valid_region(const ValidRegion &valid_region)=0
Set the valid region of the tensor.
1 channel, 1 F16 per channel
#define ARM_COMPUTE_ERROR_ON_MISMATCHING_WINDOWS(f, w)
Definition: Validate.h:183
virtual const TensorShape & tensor_shape() const =0
Size for each dimension of the tensor.
#define ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(...)
Definition: Validate.h:543
Manages all the GLES kernels compilation and caching, provides accessors for the GLES Context...
Class to describe a number of elements in each dimension.
Definition: Steps.h:40
Coordinates of an item.
Definition: Coordinates.h:37
virtual ITensorInfo * info() const =0
Interface to be implemented by the child class to return the tensor&#39;s metadata.
virtual size_t element_size() const =0
Element size in bytes calculated as data_size() * num_channels()
bool slide_window_slice_3D(Window &slice) const
Slide the passed 3D window slice.
Definition: Window.h:335
#define ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(k)
Definition: Validate.h:941
#define ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(t, c,...)
Definition: Validate.h:790
void run(const Window &window) override
Enqueue the OpenGL ES shader to process the given window.
Wrapper to configure the Khronos EGL and OpenGL ES C header.
static GCKernelLibrary & get()
Get the static instance of GCKernelLibrary.
void configure(const IGCTensor *input, IGCTensor *mask, IGCTensor *output, float ratio, bool forward)
Set the input and output of the kernel.
GCKernel create_kernel(const std::string &shader_name, const StringSet &build_options_set={}) const
Creates a kernel from the kernel library.
Container for valid region of a window.
Definition: Types.h:188
Window first_slice_window_3D() const
First 3D slice of the window.
Definition: Window.h:291
Describe a multidimensional execution window.
Definition: Window.h:39
SimpleTensor< T > slice(const SimpleTensor< T > &src, Coordinates starts, Coordinates ends)