将 saved_model.pb 转换为 model.tflite

Question

Tensorflow 版本：2.2.0

操作系统：Windows 10

我正在尝试将 saved_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(
  File "C:\Users\Mr.Ace\AppData\Roaming\Python\Python38\site-packages\tensorflow\lite\python\convert.py", line 569, in toco_convert_impl
    data = toco_convert_protos(
  File "C:\Users\Mr.Ace\AppData\Roaming\Python\Python38\site-packages\tensorflow\lite\python\convert.py", line 202, in toco_convert_protos
    raise ConverterError(str(e))
tensorflow.lite.python.convert.ConverterError: <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>

Answer 1

好的，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 tf

saved_model_dir = "C:/Data/TFOD/models/ssd_mobilenet_v2_quantized/tflite"
converter = tf.lite.TFLiteConverter.from_saved_model(
    saved_model_dir, signature_keys=['serving_default'])
converter.optimizations = [tf.lite.Optimize.DEFAULT]
converter.experimental_new_converter = True
converter.target_spec.supported_ops = [
    tf.lite.OpsSet.TFLITE_BUILTINS, tf.lite.OpsSet.SELECT_TF_OPS]
tflite_model = converter.convert()

fo = open(
    "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

Tensorflow在model/object_detection文件夹下提供了一个名为export_tflite_ssd_graph.py的python文件，可以用来将你保存的model转成tflite格式。

这是该文件的 GitHub 链接。 下载模型目录时会下载它。

如何使用它：

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 output 将在目录中
path/to/exported_model_directory（如果不存在则创建）
内容：

• tflite_graph.pbtxt
• tflite_graph.pb

如需完整使用，您可以阅读文件内的注释。