xref: /aosp_15_r20/external/tensorflow/tensorflow/core/ops/ctc_ops.cc (revision b6fb3261f9314811a0f4371741dbb8839866f948)
1 /* Copyright 2016 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 "tensorflow/core/framework/op.h"
17 #include "tensorflow/core/framework/shape_inference.h"
18 
19 namespace tensorflow {
20 
21 using shape_inference::DimensionHandle;
22 using shape_inference::InferenceContext;
23 using shape_inference::ShapeHandle;
24 
25 // CTC is Connectionist Temporal Classification.  See util/ctc/ for details.
26 
27 REGISTER_OP("CTCLoss")
28     .Input("inputs: T")
29     .Input("labels_indices: int64")
30     .Input("labels_values: int32")
31     .Input("sequence_length: int32")
32     .Attr("preprocess_collapse_repeated: bool = false")
33     .Attr("ctc_merge_repeated: bool = true")
34     .Attr("ignore_longer_outputs_than_inputs: bool = false")
35     .Output("loss: T")
36     .Output("gradient: T")
37     .Attr("T: {float, double} = DT_FLOAT")
__anon8bbafe030102(InferenceContext* c) 38     .SetShapeFn([](InferenceContext* c) {
39       ShapeHandle inputs;
40       ShapeHandle labels_indices;
41       ShapeHandle labels_values;
42       ShapeHandle sequence_length;
43 
44       TF_RETURN_IF_ERROR(c->WithRank(c->input(0), 3, &inputs));
45       TF_RETURN_IF_ERROR(c->WithRank(c->input(1), 2, &labels_indices));
46       TF_RETURN_IF_ERROR(c->WithRank(c->input(2), 1, &labels_values));
47       TF_RETURN_IF_ERROR(c->WithRank(c->input(3), 1, &sequence_length));
48 
49       DimensionHandle unused;
50       TF_RETURN_IF_ERROR(c->Merge(c->Dim(labels_indices, 0),
51                                   c->Dim(labels_values, 0), &unused));
52 
53       // Get batch size from inputs and sequence_length, and update inputs
54       // with the merged batch_size since it is returned.
55       DimensionHandle batch_size;
56       TF_RETURN_IF_ERROR(
57           c->Merge(c->Dim(inputs, 1), c->Dim(sequence_length, 0), &batch_size));
58       TF_RETURN_IF_ERROR(c->ReplaceDim(inputs, 1, batch_size, &inputs));
59 
60       c->set_output(0, c->Vector(batch_size));
61       c->set_output(1, inputs);
62       return OkStatus();
63     });
64 
65 REGISTER_OP("CTCLossV2")
66     .Input("inputs: float")
67     .Input("labels_indices: int64")
68     .Input("labels_values: int32")
69     .Input("sequence_length: int32")
70     .Attr("preprocess_collapse_repeated: bool = false")
71     .Attr("ctc_merge_repeated: bool = true")
72     .Attr("ignore_longer_outputs_than_inputs: bool = false")
73     .Output("loss: float")
74     .Output("gradient: float")
__anon8bbafe030202(InferenceContext* c) 75     .SetShapeFn([](InferenceContext* c) {
76       ShapeHandle inputs;
77       ShapeHandle labels_indices;
78       ShapeHandle labels_values;
79       ShapeHandle sequence_length;
80 
81       TF_RETURN_IF_ERROR(c->WithRank(c->input(0), 3, &inputs));
82       TF_RETURN_IF_ERROR(c->WithRank(c->input(1), 2, &labels_indices));
83       TF_RETURN_IF_ERROR(c->WithRank(c->input(2), 1, &labels_values));
84       TF_RETURN_IF_ERROR(c->WithRank(c->input(3), 1, &sequence_length));
85 
86       DimensionHandle unused;
87       TF_RETURN_IF_ERROR(c->Merge(c->Dim(labels_indices, 0),
88                                   c->Dim(labels_values, 0), &unused));
89 
90       // Get batch size from inputs and sequence_length, and update inputs
91       // with the merged batch_size since it is returned.
92       DimensionHandle batch_size;
93       TF_RETURN_IF_ERROR(
94           c->Merge(c->Dim(inputs, 1), c->Dim(sequence_length, 0), &batch_size));
95       TF_RETURN_IF_ERROR(c->ReplaceDim(inputs, 1, batch_size, &inputs));
96 
97       c->set_output(0, c->Vector(batch_size));
98       c->set_output(1, inputs);
99       return OkStatus();
100     });
101 
102 REGISTER_OP("CTCGreedyDecoder")
103     .Input("inputs: T")
104     .Input("sequence_length: int32")
105     .Attr("merge_repeated: bool = false")
106     .Attr("blank_index: int = -1")
107     .Output("decoded_indices: int64")
108     .Output("decoded_values: int64")
109     .Output("decoded_shape: int64")
110     .Output("log_probability: T")
111     .Attr("T: {float, double} = DT_FLOAT")
__anon8bbafe030302(InferenceContext* c) 112     .SetShapeFn([](InferenceContext* c) {
113       ShapeHandle inputs;
114       ShapeHandle sequence_length;
115 
116       TF_RETURN_IF_ERROR(c->WithRank(c->input(0), 3, &inputs));
117       TF_RETURN_IF_ERROR(c->WithRank(c->input(1), 1, &sequence_length));
118 
119       // Get batch size from inputs and sequence_length.
120       DimensionHandle batch_size;
121       TF_RETURN_IF_ERROR(
122           c->Merge(c->Dim(inputs, 1), c->Dim(sequence_length, 0), &batch_size));
123 
124       DimensionHandle total_decoded_outputs = c->UnknownDim();
125       c->set_output(0, c->Matrix(total_decoded_outputs, 2));
126       c->set_output(1, c->Vector(total_decoded_outputs));
127       c->set_output(2, c->Vector(2));
128       c->set_output(3, c->Matrix(batch_size, 1));
129       return OkStatus();
130     });
131 
132 REGISTER_OP("CTCBeamSearchDecoder")
133     .Input("inputs: T")
134     .Input("sequence_length: int32")
135     .Attr("beam_width: int >= 1")
136     .Attr("top_paths: int >= 1")
137     .Attr("merge_repeated: bool = true")
138     .Output("decoded_indices: top_paths * int64")
139     .Output("decoded_values: top_paths * int64")
140     .Output("decoded_shape: top_paths * int64")
141     .Output("log_probability: T")
142     .Attr("T: {float, double} = DT_FLOAT")
__anon8bbafe030402(InferenceContext* c) 143     .SetShapeFn([](InferenceContext* c) {
144       ShapeHandle inputs;
145       ShapeHandle sequence_length;
146 
147       TF_RETURN_IF_ERROR(c->WithRank(c->input(0), 3, &inputs));
148       TF_RETURN_IF_ERROR(c->WithRank(c->input(1), 1, &sequence_length));
149 
150       // Get batch size from inputs and sequence_length.
151       DimensionHandle batch_size;
152       TF_RETURN_IF_ERROR(
153           c->Merge(c->Dim(inputs, 1), c->Dim(sequence_length, 0), &batch_size));
154 
155       int32_t top_paths;
156       TF_RETURN_IF_ERROR(c->GetAttr("top_paths", &top_paths));
157 
158       // Outputs.
159       int out_idx = 0;
160       for (int i = 0; i < top_paths; ++i) {  // decoded_indices
161         c->set_output(out_idx++, c->Matrix(InferenceContext::kUnknownDim, 2));
162       }
163       for (int i = 0; i < top_paths; ++i) {  // decoded_values
164         c->set_output(out_idx++, c->Vector(InferenceContext::kUnknownDim));
165       }
166       ShapeHandle shape_v = c->Vector(2);
167       for (int i = 0; i < top_paths; ++i) {  // decoded_shape
168         c->set_output(out_idx++, shape_v);
169       }
170       c->set_output(out_idx++, c->Matrix(batch_size, top_paths));
171       return OkStatus();
172     });
173 
174 }  // namespace tensorflow
175