1# Copyright 2017 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"""Test utilities for tf.signal.""" 16 17from __future__ import absolute_import 18from __future__ import division 19from __future__ import print_function 20 21from tensorflow.core.protobuf import config_pb2 22from tensorflow.lite.python import interpreter 23from tensorflow.lite.python import lite 24from tensorflow.python.eager import def_function 25from tensorflow.python.grappler import tf_optimizer 26from tensorflow.python.training import saver 27 28 29def grappler_optimize(graph, fetches=None, config_proto=None): 30 """Tries to optimize the provided graph using grappler. 31 32 Args: 33 graph: A `tf.Graph` instance containing the graph to optimize. 34 fetches: An optional list of `Tensor`s to fetch (i.e. not optimize away). 35 Grappler uses the 'train_op' collection to look for fetches, so if not 36 provided this collection should be non-empty. 37 config_proto: An optional `tf.compat.v1.ConfigProto` to use when rewriting 38 the graph. 39 40 Returns: 41 A `tf.compat.v1.GraphDef` containing the rewritten graph. 42 """ 43 if config_proto is None: 44 config_proto = config_pb2.ConfigProto() 45 config_proto.graph_options.rewrite_options.min_graph_nodes = -1 46 if fetches is not None: 47 for fetch in fetches: 48 graph.add_to_collection('train_op', fetch) 49 metagraph = saver.export_meta_graph(graph_def=graph.as_graph_def()) 50 return tf_optimizer.OptimizeGraph(config_proto, metagraph) 51 52 53def tflite_convert(fn, input_templates): 54 """Converts the provided fn to tf.lite model. 55 56 Args: 57 fn: A callable that expects a list of inputs like input_templates that 58 returns a tensor or structure of tensors. 59 input_templates: A list of Tensors, ndarrays or TensorSpecs describing the 60 inputs that fn expects. The actual values of the Tensors or ndarrays are 61 unused. 62 63 Returns: 64 The serialized tf.lite model. 65 """ 66 fn = def_function.function(fn) 67 concrete_func = fn.get_concrete_function(*input_templates) 68 converter = lite.TFLiteConverterV2([concrete_func]) 69 return converter.convert() 70 71 72def evaluate_tflite_model(tflite_model, input_ndarrays): 73 """Evaluates the provided tf.lite model with the given input ndarrays. 74 75 Args: 76 tflite_model: bytes. The serialized tf.lite model. 77 input_ndarrays: A list of NumPy arrays to feed as input to the model. 78 79 Returns: 80 A list of ndarrays produced by the model. 81 82 Raises: 83 ValueError: If the number of input arrays does not match the number of 84 inputs the model expects. 85 """ 86 the_interpreter = interpreter.Interpreter(model_content=tflite_model) 87 the_interpreter.allocate_tensors() 88 89 input_details = the_interpreter.get_input_details() 90 output_details = the_interpreter.get_output_details() 91 92 if len(input_details) != len(input_ndarrays): 93 raise ValueError('Wrong number of inputs: provided=%s, ' 94 'input_details=%s output_details=%s' % ( 95 input_ndarrays, input_details, output_details)) 96 for input_tensor, data in zip(input_details, input_ndarrays): 97 the_interpreter.set_tensor(input_tensor['index'], data) 98 the_interpreter.invoke() 99 return [the_interpreter.get_tensor(details['index']) 100 for details in output_details] 101