将 save_model.pb 转换为 model.tflite

Question

张量流版本：2.2.0

操作系统：Windows 10

我正在尝试将 saving_model.pb 转换为 tflite 文件。

这是我正在运行的代码：

import tensorflow as tf

# Convert

converter = tf.lite.TFLiteConverter.from_saved_model(saved_model_dir='C:\Data\TFOD\models\ssd_mobilenet_v2_quantized')

tflite_model = converter.convert()

fo = open("model.tflite", "wb")

fo.write(tflite_model)

fo.close

此代码在转换时出错：

  File "C:\Users\Mr.Ace\AppData\Roaming\Python\Python38\site-packages\tensorflow\lite\python\convert.py", line 196, in toco_convert_protos

    model_str = wrap_toco.wrapped_toco_convert(model_flags_str,

  File "C:\Users\Mr.Ace\AppData\Roaming\Python\Python38\site-packages\tensorflow\lite\python\wrap_toco.py", line 32, in wrapped_toco_convert

    return _pywrap_toco_api.TocoConvert(

Exception: <unknown>:0: error: loc("Func/StatefulPartitionedCall/input/_0"): requires all operands and results to have compatible element types

<unknown>:0: note: loc("Func/StatefulPartitionedCall/input/_0"): see current operation: %1 = "tf.Identity"(%arg0) {device = ""} : (tensor<1x?x?x3x!tf.quint8>) -> tensor<1x?x?x3xui8>

During handling of the above exception, another exception occurred:

Traceback (most recent call last):

  File "c:/Data/TFOD/convert.py", line 13, in <module>

    tflite_model = converter.convert()

  File "C:\Users\Mr.Ace\AppData\Roaming\Python\Python38\site-packages\tensorflow\lite\python\lite.py", line 1076, in convert

    return super(TFLiteConverterV2, self).convert()

  File "C:\Users\Mr.Ace\AppData\Roaming\Python\Python38\site-packages\tensorflow\lite\python\lite.py", line 899, in convert

    return super(TFLiteFrozenGraphConverterV2,

  File "C:\Users\Mr.Ace\AppData\Roaming\Python\Python38\site-packages\tensorflow\lite\python\lite.py", line 629, in convert

    result = _toco_convert_impl(

Answer 1

好吧，我终于解决了！我所做的是使用 tf-nightly 并使用以下 Python 脚本：import tensorflow as tfsaved_model_dir = "C:/Data/TFOD/models/ssd_mobilenet_v2_quantized/tflite"converter = tf.lite.TFLiteConverter.from_saved_model(&nbsp; &nbsp; saved_model_dir, signature_keys=['serving_default'])converter.optimizations = [tf.lite.Optimize.DEFAULT]converter.experimental_new_converter = Trueconverter.target_spec.supported_ops = [&nbsp; &nbsp; tf.lite.OpsSet.TFLITE_BUILTINS, tf.lite.OpsSet.SELECT_TF_OPS]tflite_model = converter.convert()fo = open(&nbsp; &nbsp; "C:/Data/TFOD/models/ssd_mobilenet_v2_quantized/tflite/model.tflite", "wb")fo.write(tflite_model)fo.close这解决了问题，您可以转换为 .tflite

Answer 2

export_tflite_ssd_graph.pyTensorflow在该文件夹中提供了一个名为 python 文件model/object_detection，可用于将保存的模型转换为 tflite 格式。python object_detection/export_tflite_ssd_graph.py \     --pipeline_config_path path/to/ssd_mobilenet.config \     --trained_checkpoint_prefix path/to/model.ckpt \     --output_directory path/to/exported_model_directory预期的输出将位于目录path/to/exported_model_directory （如果不存在则创建），其内容为：tflite_graph.pbtxttflite_graph.pb如需完整使用，您可以阅读文件内的注释。