Lukas Zilka | 21d8c98 | 2018-01-24 11:11:20 +0100 | [diff] [blame] | 1 | /* |
| 2 | * Copyright (C) 2017 The Android Open Source Project |
| 3 | * |
| 4 | * Licensed under the Apache License, Version 2.0 (the "License"); |
| 5 | * you may not use this file except in compliance with the License. |
| 6 | * You may obtain a copy of the License at |
| 7 | * |
| 8 | * http://www.apache.org/licenses/LICENSE-2.0 |
| 9 | * |
| 10 | * Unless required by applicable law or agreed to in writing, software |
| 11 | * distributed under the License is distributed on an "AS IS" BASIS, |
| 12 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 13 | * See the License for the specific language governing permissions and |
| 14 | * limitations under the License. |
| 15 | */ |
| 16 | |
| 17 | #ifndef KNOWLEDGE_CEREBRA_SENSE_TEXT_CLASSIFIER_LIB2_CACHED_FEATURES_H_ |
| 18 | #define KNOWLEDGE_CEREBRA_SENSE_TEXT_CLASSIFIER_LIB2_CACHED_FEATURES_H_ |
| 19 | |
| 20 | #include <memory> |
| 21 | #include <vector> |
| 22 | |
| 23 | #include "model-executor.h" |
| 24 | #include "model_generated.h" |
| 25 | #include "types.h" |
| 26 | |
| 27 | namespace libtextclassifier2 { |
| 28 | |
| 29 | // Holds state for extracting features across multiple calls and reusing them. |
| 30 | // Assumes that features for each Token are independent. |
| 31 | class CachedFeatures { |
| 32 | public: |
| 33 | CachedFeatures( |
| 34 | const TokenSpan& extraction_span, |
| 35 | const std::vector<std::vector<int>>& sparse_features, |
| 36 | const std::vector<std::vector<float>>& dense_features, |
| 37 | const std::vector<int>& padding_sparse_features, |
| 38 | const std::vector<float>& padding_dense_features, |
| 39 | const FeatureProcessorOptions_::BoundsSensitiveFeatures* config, |
| 40 | EmbeddingExecutor* embedding_executor, int feature_vector_size); |
| 41 | |
| 42 | // Gets a vector of features for the given token span. |
| 43 | std::vector<float> Get(TokenSpan selected_span) const; |
| 44 | |
| 45 | private: |
| 46 | // Appends token features to the output. The intended_span specifies which |
| 47 | // tokens' features should be used in principle. The read_mask_span restricts |
| 48 | // which tokens are actually read. For tokens outside of the read_mask_span, |
| 49 | // padding tokens are used instead. |
| 50 | void AppendFeatures(const TokenSpan& intended_span, |
| 51 | const TokenSpan& read_mask_span, |
| 52 | std::vector<float>* output_features) const; |
| 53 | |
| 54 | // Appends features of one padding token to the output. |
| 55 | void AppendPaddingFeatures(std::vector<float>* output_features) const; |
| 56 | |
| 57 | // Appends the features of tokens from the given span to the output. The |
| 58 | // features are summed so that the appended features have the size |
| 59 | // corresponding to one token. |
| 60 | void AppendSummedFeatures(const TokenSpan& summing_span, |
| 61 | std::vector<float>* output_features) const; |
| 62 | |
| 63 | int NumFeaturesPerToken() const; |
| 64 | |
| 65 | const TokenSpan extraction_span_; |
| 66 | const FeatureProcessorOptions_::BoundsSensitiveFeatures* config_; |
| 67 | int output_features_size_; |
| 68 | std::vector<float> features_; |
| 69 | std::vector<float> padding_features_; |
| 70 | }; |
| 71 | |
| 72 | } // namespace libtextclassifier2 |
| 73 | |
| 74 | #endif // KNOWLEDGE_CEREBRA_SENSE_TEXT_CLASSIFIER_LIB2_CACHED_FEATURES_H_ |