def get_attention_per_path(self, source_strings, path_strings, target_strings, attention_weights): attention_weights = np.squeeze(attention_weights) # (max_contexts, ) attention_per_context = {} for source, path, target, weight in zip(source_strings, path_strings, target_strings, attention_weights): string_triplet = ( common.binary_to_string(source), common.binary_to_string(path), common.binary_to_string(target)) attention_per_context[string_triplet] = weight return attention_per_context
def _get_embed_per_context( self, path_source_strings: Iterable[str], path_strings: Iterable[str], path_target_strings: Iterable[str], embeddings: Iterable[np.ndarray]) -> Dict[Tuple[str, str, str], np.ndarray]: embed_per_context: Dict[Tuple[str, str, str], np.ndarray] = {} # iterate over contexts for path_source, path, path_target, embed in \ zip(path_source_strings, path_strings, path_target_strings, embeddings): string_context_triplet = (common.binary_to_string(path_source), common.binary_to_string(path), common.binary_to_string(path_target)) embed_per_context[string_context_triplet] = embed return embed_per_context
def _get_attention_weight_per_context( self, path_source_strings: Iterable[str], path_strings: Iterable[str], path_target_strings: Iterable[str], attention_weights: Iterable[float]) -> Dict[Tuple[str, str, str], float]: attention_weights = np.squeeze(attention_weights, axis=-1) # (max_contexts, ) attention_per_context: Dict[Tuple[str, str, str], float] = {} # shape of path_source_strings, path_strings, path_target_strings, attention_weights is (max_contexts, ) # iterate over contexts for path_source, path, path_target, weight in \ zip(path_source_strings, path_strings, path_target_strings, attention_weights): string_context_triplet = (common.binary_to_string(path_source), common.binary_to_string(path), common.binary_to_string(path_target)) attention_per_context[string_context_triplet] = weight return attention_per_context
def predict(self, predict_data_rows: Iterable[str]) -> List[ModelPredictionResults]: predict_input_reader = self._create_data_reader(estimator_action=EstimatorAction.Predict) input_iterator = predict_input_reader.process_and_iterate_input_from_data_lines(predict_data_rows) all_model_prediction_results = [] for input_row in input_iterator: # perform the actual prediction and get raw results. input_for_predict = input_row[0][:4] # we want only the relevant input vectors (w.o. the targets). prediction_results = self.keras_model_predict_function(input_for_predict) # make `input_row` and `prediction_results` easy to read (by accessing named fields). prediction_results = KerasPredictionModelOutput( *common.squeeze_single_batch_dimension_for_np_arrays(prediction_results)) input_row = _KerasModelInputTensorsFormer( estimator_action=EstimatorAction.Predict).from_model_input_form(input_row) input_row = ReaderInputTensors(*common.squeeze_single_batch_dimension_for_np_arrays(input_row)) # calculate the attention weight for each context attention_per_context = self._get_attention_weight_per_context( path_source_strings=input_row.path_source_token_strings, path_strings=input_row.path_strings, path_target_strings=input_row.path_target_token_strings, attention_weights=prediction_results.attention_weights ) # store the calculated prediction results in the wanted format. model_prediction_results = ModelPredictionResults( original_name=common.binary_to_string(input_row.target_string.item()), topk_predicted_words=common.binary_to_string_list(prediction_results.topk_predicted_words), topk_predicted_words_scores=prediction_results.topk_predicted_words_scores, attention_per_context=attention_per_context, code_vector=prediction_results.code_vectors) all_model_prediction_results.append(model_prediction_results) return all_model_prediction_results
def _get_embed_per_path(self, path_strings: Iterable[str], embeddings: Iterable[np.ndarray]) -> Dict[str, np.ndarray]: embed_per_path: Dict[str, np.ndarray] = {} # iterate over contexts for path, embed in zip(path_strings, embeddings): embed_per_path[common.binary_to_string(path)] = embed return embed_per_path
def predict(self, predict_data_lines: Iterable[str]) -> List[ModelPredictionResults]: if self.predict_reader is None: self.predict_reader = PathContextReader(vocabs=self.vocabs, model_input_tensors_former=_TFEvaluateModelInputTensorsFormer(), config=self.config, estimator_action=EstimatorAction.Predict) self.predict_placeholder = tf.compat.v1.placeholder(tf.string) reader_output = self.predict_reader.process_input_row(self.predict_placeholder) self.predict_top_words_op, self.predict_top_values_op, self.predict_original_names_op, \ self.attention_weights_op, self.predict_source_string, self.predict_path_string, \ self.predict_path_target_string, self.predict_code_vectors = \ self._build_tf_test_graph(reader_output, normalize_scores=True) self._initialize_session_variables() self.saver = tf.compat.v1.train.Saver() self._load_inner_model(sess=self.sess) prediction_results: List[ModelPredictionResults] = [] for line in predict_data_lines: batch_top_words, batch_top_scores, batch_original_name, batch_attention_weights, batch_path_source_strings,\ batch_path_strings, batch_path_target_strings, batch_code_vectors = self.sess.run( [self.predict_top_words_op, self.predict_top_values_op, self.predict_original_names_op, self.attention_weights_op, self.predict_source_string, self.predict_path_string, self.predict_path_target_string, self.predict_code_vectors], feed_dict={self.predict_placeholder: line}) # shapes: # batch_top_words, top_scores: (batch, top_k) # batch_original_name: (batch, ) # batch_attention_weights: (batch, max_context, 1) # batch_path_source_strings, batch_path_strings, batch_path_target_strings: (batch, max_context) # batch_code_vectors: (batch, code_vector_size) # remove first axis: (batch=1, ...) assert all(tensor.shape[0] == 1 for tensor in (batch_top_words, batch_top_scores, batch_original_name, batch_attention_weights, batch_path_source_strings, batch_path_strings, batch_path_target_strings, batch_code_vectors)) top_words = np.squeeze(batch_top_words, axis=0) top_scores = np.squeeze(batch_top_scores, axis=0) original_name = batch_original_name[0] attention_weights = np.squeeze(batch_attention_weights, axis=0) path_source_strings = np.squeeze(batch_path_source_strings, axis=0) path_strings = np.squeeze(batch_path_strings, axis=0) path_target_strings = np.squeeze(batch_path_target_strings, axis=0) code_vectors = np.squeeze(batch_code_vectors, axis=0) top_words = common.binary_to_string_list(top_words) original_name = common.binary_to_string(original_name) attention_per_context = self._get_attention_weight_per_context( path_source_strings, path_strings, path_target_strings, attention_weights) prediction_results.append(ModelPredictionResults( original_name=original_name, topk_predicted_words=top_words, topk_predicted_words_scores=top_scores, attention_per_context=attention_per_context, code_vector=(code_vectors if self.config.EXPORT_CODE_VECTORS else None) )) return prediction_results