1 | /* Copyright 2021 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 | #include <stdint.h> |
17 | |
18 | #include <algorithm> |
19 | |
20 | #include "tensorflow/lite/c/builtin_op_data.h" |
21 | #include "tensorflow/lite/c/common.h" |
22 | #include "tensorflow/lite/kernels/internal/tensor.h" |
23 | #include "tensorflow/lite/kernels/internal/tensor_ctypes.h" |
24 | #include "tensorflow/lite/kernels/kernel_util.h" |
25 | |
26 | namespace tflite { |
27 | namespace ops { |
28 | namespace builtin { |
29 | namespace bucketize { |
30 | namespace { |
31 | |
32 | constexpr int kInputTensor = 0; |
33 | constexpr int kOutputTensor = 0; |
34 | |
35 | struct OpData { |
36 | // boundaries array is owned by the buffer housing TfLiteBucketizeParams. |
37 | const float* boundaries; |
38 | int num_boundaries; |
39 | }; |
40 | |
41 | void* Init(TfLiteContext* context, const char* buffer, size_t length) { |
42 | auto* op_data = new OpData(); |
43 | const auto* params = reinterpret_cast<const TfLiteBucketizeParams*>(buffer); |
44 | |
45 | op_data->boundaries = params->boundaries; |
46 | op_data->num_boundaries = params->num_boundaries; |
47 | return op_data; |
48 | } |
49 | |
50 | void Free(TfLiteContext* context, void* buffer) { |
51 | delete reinterpret_cast<OpData*>(buffer); |
52 | } |
53 | |
54 | TfLiteStatus Prepare(TfLiteContext* context, TfLiteNode* node) { |
55 | TF_LITE_ENSURE_EQ(context, NumInputs(node), 1); |
56 | TF_LITE_ENSURE_EQ(context, NumOutputs(node), 1); |
57 | OpData* opdata = reinterpret_cast<OpData*>(node->user_data); |
58 | if (!std::is_sorted(opdata->boundaries, |
59 | opdata->boundaries + opdata->num_boundaries)) { |
60 | TF_LITE_KERNEL_LOG(context, "Expected sorted boundaries" ); |
61 | return kTfLiteError; |
62 | } |
63 | |
64 | const TfLiteTensor* input; |
65 | TF_LITE_ENSURE_OK(context, GetInputSafe(context, node, kInputTensor, &input)); |
66 | |
67 | if (input->type != kTfLiteInt32 && input->type != kTfLiteFloat32 && |
68 | input->type != kTfLiteInt64 && input->type != kTfLiteFloat64) { |
69 | TF_LITE_KERNEL_LOG(context, "Type '%s' is not supported by bucketize." , |
70 | TfLiteTypeGetName(input->type)); |
71 | return kTfLiteError; |
72 | } |
73 | |
74 | TfLiteTensor* output; |
75 | TF_LITE_ENSURE_OK(context, |
76 | GetOutputSafe(context, node, kOutputTensor, &output)); |
77 | output->type = kTfLiteInt32; |
78 | |
79 | TfLiteIntArray* output_shape = TfLiteIntArrayCopy(input->dims); |
80 | return context->ResizeTensor(context, output, output_shape); |
81 | } |
82 | |
83 | template <typename T> |
84 | inline void Bucketize(const RuntimeShape& input_shape, const T* input_data, |
85 | const float* boundaries, int num_boundaries, |
86 | const RuntimeShape& output_shape, int32_t* output_data) { |
87 | const int flat_size = MatchingFlatSize(input_shape, output_shape); |
88 | |
89 | for (int i = 0; i < flat_size; i++) { |
90 | auto first_bigger_it = std::upper_bound( |
91 | boundaries, boundaries + num_boundaries, input_data[i]); |
92 | output_data[i] = first_bigger_it - boundaries; |
93 | } |
94 | } |
95 | |
96 | template <typename T> |
97 | TfLiteStatus BucketizeImpl(TfLiteContext* context, TfLiteNode* node) { |
98 | const TfLiteTensor* input; |
99 | TF_LITE_ENSURE_OK(context, GetInputSafe(context, node, kInputTensor, &input)); |
100 | OpData* opdata = reinterpret_cast<OpData*>(node->user_data); |
101 | TfLiteTensor* output; |
102 | TF_LITE_ENSURE_OK(context, |
103 | GetOutputSafe(context, node, kOutputTensor, &output)); |
104 | TF_LITE_ENSURE_TYPES_EQ(context, output->type, kTfLiteInt32); |
105 | |
106 | Bucketize<T>(GetTensorShape(input), GetTensorData<T>(input), |
107 | opdata->boundaries, opdata->num_boundaries, |
108 | GetTensorShape(output), GetTensorData<int32_t>(output)); |
109 | |
110 | return kTfLiteOk; |
111 | } |
112 | |
113 | TfLiteStatus Eval(TfLiteContext* context, TfLiteNode* node) { |
114 | const TfLiteTensor* input; |
115 | TF_LITE_ENSURE_OK(context, GetInputSafe(context, node, kInputTensor, &input)); |
116 | |
117 | switch (input->type) { |
118 | case kTfLiteFloat32: { |
119 | return BucketizeImpl<float>(context, node); |
120 | } |
121 | case kTfLiteFloat64: { |
122 | return BucketizeImpl<double>(context, node); |
123 | } |
124 | case kTfLiteInt32: { |
125 | return BucketizeImpl<int32_t>(context, node); |
126 | } |
127 | case kTfLiteInt64: { |
128 | return BucketizeImpl<int64_t>(context, node); |
129 | } |
130 | default: { |
131 | TF_LITE_KERNEL_LOG(context, "Type '%s' is not supported by bucketize." , |
132 | TfLiteTypeGetName(input->type)); |
133 | return kTfLiteError; |
134 | } |
135 | } |
136 | } |
137 | |
138 | } // namespace |
139 | } // namespace bucketize |
140 | |
141 | TfLiteRegistration* Register_BUCKETIZE() { |
142 | static TfLiteRegistration r = {bucketize::Init, bucketize::Free, |
143 | bucketize::Prepare, bucketize::Eval}; |
144 | return &r; |
145 | } |
146 | |
147 | } // namespace builtin |
148 | } // namespace ops |
149 | } // namespace tflite |
150 | |