|
| 1 | +#!/usr/bin/env python3 |
| 2 | +# Copyright (c) 2021 Arm Limited. |
| 3 | +# |
| 4 | +# SPDX-License-Identifier: MIT |
| 5 | +# |
| 6 | +# Permission is hereby granted, free of charge, to any person obtaining a copy |
| 7 | +# of this software and associated documentation files (the "Software"), to |
| 8 | +# deal in the Software without restriction, including without limitation the |
| 9 | +# rights to use, copy, modify, merge, publish, distribute, sublicense, and/or |
| 10 | +# sell copies of the Software, and to permit persons to whom the Software is |
| 11 | +# furnished to do so, subject to the following conditions: |
| 12 | +# |
| 13 | +# The above copyright notice and this permission notice shall be included in all |
| 14 | +# copies or substantial portions of the Software. |
| 15 | +# |
| 16 | +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
| 17 | +# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 18 | +# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE |
| 19 | +# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
| 20 | +# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
| 21 | +# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE |
| 22 | +# SOFTWARE. |
| 23 | +import json |
| 24 | +import logging |
| 25 | +import os |
| 26 | +import sys |
| 27 | +from argparse import ArgumentParser |
| 28 | + |
| 29 | +import tflite |
| 30 | + |
| 31 | +sys.path.append(os.path.dirname(os.path.abspath(__file__)) + "/../") |
| 32 | + |
| 33 | +from utils.model_identification import identify_model_type |
| 34 | +from utils.tflite_helpers import tflite_op2acl, tflite_typecode2name |
| 35 | + |
| 36 | +SUPPORTED_MODEL_TYPES = ["tflite"] |
| 37 | +logger = logging.getLogger("report_model_ops") |
| 38 | + |
| 39 | + |
| 40 | +def get_ops_from_tflite_graph(model): |
| 41 | + """ |
| 42 | + Helper function that extract operator related meta-data from a TfLite model |
| 43 | +
|
| 44 | + Parameters |
| 45 | + ---------- |
| 46 | + model: str |
| 47 | + Respective TfLite model to analyse |
| 48 | +
|
| 49 | + Returns |
| 50 | + ---------- |
| 51 | + supported_ops, unsupported_ops, data_types: tuple |
| 52 | + A tuple with the sets of unique operator types and data-types that are present in the model |
| 53 | + """ |
| 54 | + |
| 55 | + logger.debug(f"Analysing TfLite mode '{model}'!") |
| 56 | + |
| 57 | + with open(model, "rb") as f: |
| 58 | + buf = f.read() |
| 59 | + model = tflite.Model.GetRootAsModel(buf, 0) |
| 60 | + |
| 61 | + # Extract unique operators |
| 62 | + nr_unique_ops = model.OperatorCodesLength() |
| 63 | + unique_ops = {tflite.opcode2name(model.OperatorCodes(op_id).BuiltinCode()) for op_id in range(0, nr_unique_ops)} |
| 64 | + |
| 65 | + # Extract IO data-types |
| 66 | + data_types = set() |
| 67 | + for subgraph_id in range(0, model.SubgraphsLength()): |
| 68 | + subgraph = model.Subgraphs(subgraph_id) |
| 69 | + for tensor_id in range(0, subgraph.TensorsLength()): |
| 70 | + data_types.add(tflite_typecode2name(subgraph.Tensors(tensor_id).Type())) |
| 71 | + |
| 72 | + # Perform mapping between TfLite ops to ComputeLibrary ones |
| 73 | + supported_ops = set() |
| 74 | + unsupported_ops = set() |
| 75 | + for top in unique_ops: |
| 76 | + try: |
| 77 | + supported_ops.add(tflite_op2acl(top)) |
| 78 | + except: |
| 79 | + unsupported_ops.add(top) |
| 80 | + logger.warning(f"Operator {top} has not ComputeLibrary mapping") |
| 81 | + |
| 82 | + return (supported_ops, unsupported_ops, data_types) |
| 83 | + |
| 84 | + |
| 85 | +def extract_model_meta(model, model_type): |
| 86 | + """ |
| 87 | + Function that calls the appropriate model parser to extract model related meta-data |
| 88 | + Supported parsers: TfLite |
| 89 | +
|
| 90 | + Parameters |
| 91 | + ---------- |
| 92 | + model: str |
| 93 | + Path to model that we want to analyze |
| 94 | + model_type: |
| 95 | + type of the model |
| 96 | +
|
| 97 | + Returns |
| 98 | + ---------- |
| 99 | + ops, data_types: (tuple) |
| 100 | + A tuple with the list of unique operator types and data-types that are present in the model |
| 101 | + """ |
| 102 | + |
| 103 | + if model_type == "tflite": |
| 104 | + return get_ops_from_tflite_graph(model) |
| 105 | + else: |
| 106 | + logger.warning(f"Model type '{model_type}' is unsupported!") |
| 107 | + return () |
| 108 | + |
| 109 | + |
| 110 | +def generate_build_config(ops, data_types): |
| 111 | + """ |
| 112 | + Function that generates a compatible ComputeLibrary operator-based build configuration |
| 113 | +
|
| 114 | + Parameters |
| 115 | + ---------- |
| 116 | + ops: set |
| 117 | + Set with the operators to add in the build configuration |
| 118 | + data_types: |
| 119 | + Set with the data types to add in the build configuration |
| 120 | +
|
| 121 | + Returns |
| 122 | + ---------- |
| 123 | + config_data: dict |
| 124 | + Dictionary compatible with ComputeLibrary |
| 125 | + """ |
| 126 | + config_data = {} |
| 127 | + config_data["operators"] = list(ops) |
| 128 | + config_data["data_types"] = list(data_types) |
| 129 | + |
| 130 | + return config_data |
| 131 | + |
| 132 | + |
| 133 | +if __name__ == "__main__": |
| 134 | + parser = ArgumentParser( |
| 135 | + description="""Report map of operations in a list of models. |
| 136 | + The script consumes deep learning models and reports the type of operations and data-types used |
| 137 | + Supported model types: TfLite """ |
| 138 | + ) |
| 139 | + |
| 140 | + parser.add_argument( |
| 141 | + "-m", |
| 142 | + "--models", |
| 143 | + nargs="+", |
| 144 | + required=True, |
| 145 | + type=str, |
| 146 | + help=f"List of models; supported model types: {SUPPORTED_MODEL_TYPES}", |
| 147 | + ) |
| 148 | + parser.add_argument("-D", "--debug", action="store_true", help="Enable script debugging output") |
| 149 | + parser.add_argument( |
| 150 | + "-c", |
| 151 | + "--config", |
| 152 | + type=str, |
| 153 | + help="JSON configuration file used that can be used for custom ComputeLibrary builds", |
| 154 | + ) |
| 155 | + args = parser.parse_args() |
| 156 | + |
| 157 | + # Setup Logger |
| 158 | + logging_level = logging.INFO |
| 159 | + if args.debug: |
| 160 | + logging_level = logging.DEBUG |
| 161 | + logging.basicConfig(level=logging_level) |
| 162 | + |
| 163 | + # Extract operator mapping |
| 164 | + final_supported_ops = set() |
| 165 | + final_unsupported_ops = set() |
| 166 | + final_dts = set() |
| 167 | + for model in args.models: |
| 168 | + logger.debug(f"Starting analyzing {model} model") |
| 169 | + |
| 170 | + model_type = identify_model_type(model) |
| 171 | + supported_model_ops, unsupported_mode_ops, model_dts = extract_model_meta(model, model_type) |
| 172 | + final_supported_ops.update(supported_model_ops) |
| 173 | + final_unsupported_ops.update(unsupported_mode_ops) |
| 174 | + final_dts.update(model_dts) |
| 175 | + |
| 176 | + logger.info("=== Supported Operators") |
| 177 | + logger.info(final_supported_ops) |
| 178 | + logger.info("=== Unsupported Operators") |
| 179 | + logger.info(final_unsupported_ops) |
| 180 | + logger.info("=== Data Types") |
| 181 | + logger.info(final_dts) |
| 182 | + |
| 183 | + # Generate json file |
| 184 | + if args.config: |
| 185 | + logger.debug("Generating JSON build configuration file") |
| 186 | + config_data = generate_build_config(final_supported_ops, final_dts) |
| 187 | + with open(args.config, "w") as f: |
| 188 | + json.dump(config_data, f) |
0 commit comments