Tensorflow Numpy 函数可以转换为 tensorflow lite 模型吗？

Question

我有一个 python 函数，它接受一个 Numpy 数组并返回另一个 Numpy 数组，我们称之为getFeatures， 函数输出是将作为输入传递给 Tensorflow lite 模型进行分类的特征，因为 Tensorflow 支持 numpy_function，它包装了一个 python 函数并将其用作 TensorFlow op 我想使用 numpy_function 作为 tflite 转换 getFeatures模型，所以我不必为 android 开发getFeatures 来提取特征。

问题是我在使用这段代码时遇到了一些错误：

from __future__ import absolute_import, division, print_function, unicode_literals
import traceback
import textwrap
import tensorflow as tf

def getFeatures(dataArray): 
    x = features(dataArray, 1600)
    return x 

@tf.function
def tf_function(dataArray): 
    y = tf.numpy_function(getFeatures, [dataArray], tf.double)
    return y

root = tf.train.Checkpoint()
root.f = tf_function
input_data = np.array([1,3])
concrete_func = root.f.get_concrete_function(input_data)
converterC = tf.lite.TFLiteConverter.from_concrete_functions([concrete_func])
tfliteC = converterC.convert()
open("FeaturesModel/tfliteFunc", "wb").write(tfliteC)

这是错误：

2020-01-15 16:29:09.823958: I tensorflow/core/grappler/devices.cc:60] Number of eligible GPUs (core count >= 8, compute capability >= 0.0): 0 (Note: TensorFlow was not compiled with CUDA support)
2020-01-15 16:29:09.879081: I tensorflow/core/grappler/clusters/single_machine.cc:356] Starting new session
2020-01-15 16:29:09.882306: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
2020-01-15 16:29:09.889356: I tensorflow/core/grappler/optimizers/meta_optimizer.cc:716] Optimization results for grappler item: graph_to_optimize
2020-01-15 16:29:09.891628: I tensorflow/core/grappler/optimizers/meta_optimizer.cc:718]   function_optimizer: function_optimizer did nothing. time = 0.003ms.
2020-01-15 16:29:09.901985: I tensorflow/core/grappler/optimizers/meta_optimizer.cc:718]   function_optimizer: function_optimizer did nothing. time = 0ms.
2020-01-15 16:29:09.916149: I tensorflow/core/grappler/devices.cc:60] Number of eligible GPUs (core count >= 8, compute capability >= 0.0): 0 (Note: TensorFlow was not compiled with CUDA support)
2020-01-15 16:29:09.920390: I tensorflow/core/grappler/clusters/single_machine.cc:356] Starting new session
Traceback (most recent call last):
  File "mlFeaturesToTFlite.py", line 31, in <module>
    tfliteC = converterC.convert()
  File "C:\Users\Lafi\AppData\Local\Programs\Python\Python36\lib\site-packages\tensorflow_core\lite\python\lite.py", line 446, in convert
    **converter_kwargs)
  File "C:\Users\Lafi\AppData\Local\Programs\Python\Python36\lib\site-packages\tensorflow_core\lite\python\convert.py", line 449, in toco_convert_impl
    enable_mlir_converter=enable_mlir_converter)
  File "C:\Users\Lafi\AppData\Local\Programs\Python\Python36\lib\site-packages\tensorflow_core\lite\python\convert.py", line 200, in toco_convert_protos
    raise ConverterError("See console for info.\n%s\n%s\n" % (stdout, stderr))
tensorflow.lite.python.convert.ConverterError: See console for info.
2020-01-15 16:29:11.883141: I tensorflow/lite/toco/import_tensorflow.cc:659] Converting unsupported operation: PyFunc
2020-01-15 16:29:11.883329: F tensorflow/lite/toco/import_tensorflow.cc:114] Check failed: attr.value_case() == AttrValue::kType (1 vs. 6)
Fatal Python error: Aborted

Current thread 0x00002180 (most recent call first):
  File "c:\users\lafi\appdata\local\programs\python\python36\lib\site-packages\tensorflow_core\lite\toco\python\toco_from_protos.py", line 52 in execute
  File "c:\users\lafi\appdata\local\programs\python\python36\lib\site-packages\absl\app.py", line 250 in _run_main
  File "c:\users\lafi\appdata\local\programs\python\python36\lib\site-packages\absl\app.py", line 299 in run
  File "c:\users\lafi\appdata\local\programs\python\python36\lib\site-packages\tensorflow_core\python\platform\app.py", line 40 in run
  File "c:\users\lafi\appdata\local\programs\python\python36\lib\site-packages\tensorflow_core\lite\toco\python\toco_from_protos.py", line 89 in main
  File "C:\Users\Lafi\AppData\Local\Programs\Python\Python36\Scripts\toco_from_protos.exe\__main__.py", line 7 in <module>
  File "c:\users\lafi\appdata\local\programs\python\python36\lib\runpy.py", line 85 in _run_code
  File "c:\users\lafi\appdata\local\programs\python\python36\lib\runpy.py", line 193 in _run_module_as_main

Answer 1

简短的回答是否。 TFLite 中没有 Numpy 数组。