Блог

Env: Tensorflow2.3.0 python3.6

Я пытаюсь настроить слой для процесса обучения для увеличения изображения. Вот мой код:

 class RandomLight(layers.Layer):
def __init__(self, factor=0.2):
    super(RandomLight,self).__init__()
    self.factor = factor

def call(self, input, training=None):
    return tf.cond(training,
                  lambda: tf.clip_by_value(tf.image.random_brightness(input,self.factor),0,1),
                  lambda: input)
  

и когда я собираюсь поместить его в сеть:

 import tensorflow.keras as keras
import tensorflow.keras.layers as layers
from tensorflow.keras.applications import VGG16

inputs = keras.Input(shape=(224,224,3))
vgg16 = VGG16(include_top=False, weights='imagenet',input_shape=(224,224,3))
data_augmentation = keras.Sequential(
[
    layers.experimental.preprocessing.RandomRotation(0.25),
    layers.experimental.preprocessing.RandomFlip(),
    RandomLight()
])
i1 = data_augmentation(inputs)
bn = layers.BatchNormalization()(i1)
x = vgg16(bn)
flat_out = layers.Flatten()(x)
h1 = layers.Dense(1024,activation='relu',name='fc1')(flat_out)
h2 = layers.Dropout(0.5)(h1)
h3 = layers.Dense(32,activation='relu',name='fc2')(h2)
h4 = layers.Dropout(0.5)(h3)
new_out = layers.Dense(1,activation='sigmoid',name='prediction')(h4)
vgg_ft = keras.Model(inputs,new_out)
  

Кажется, что возникает ошибка «обучение = Нет»

 ValueError                                Traceback (most recent call last)
<ipython-input-290-966a2fabc71b> in <module>()
----> 1 inputs = data_augmentation(inputs)
  2 inputs = randomLight(inputs)
  3 bn = layers.BatchNormalization()(inputs)
  4 x = vgg16(bn)
  5 flat_out = layers.Flatten()(x)

F:Anaconda3envstflibsite-packagestensorflowpythonkerasenginebase_layer.py in __call__(self, *args, **kwargs)
    924     if _in_functional_construction_mode(self, inputs, args, kwargs, input_list):
    925       return self._functional_construction_call(inputs, args, kwargs,
--> 926                                                 input_list)
    927 
    928     # Maintains info about the `Layer.call` stack.

F:Anaconda3envstflibsite-packagestensorflowpythonkerasenginebase_layer.py in _functional_construction_call(self, inputs, args, kwargs, input_list)
   1115           try:
   1116             with ops.enable_auto_cast_variables(self._compute_dtype_object):
-> 1117               outputs = call_fn(cast_inputs, *args, **kwargs)
   1118 
   1119           except errors.OperatorNotAllowedInGraphError as e:

F:Anaconda3envstflibsite-packagestensorflowpythonautographimplapi.py in wrapper(*args, **kwargs)
    256       except Exception as e:  # pylint:disable=broad-except
    257         if hasattr(e, 'ag_error_metadata'):
--> 258           raise e.ag_error_metadata.to_exception(e)
    259         else:
    260           raise

ValueError: in user code:
<ipython-input-278-87ec004f05b3>:11 call  *
    lambda: input)
F:Anaconda3envstflibsite-packagestensorflowpythonutildispatch.py:201 wrapper  **
    return target(*args, **kwargs)
F:Anaconda3envstflibsite-packagestensorflowpythonopscontrol_flow_ops.py:1396 cond_for_tf_v2
    return cond(pred, true_fn=true_fn, false_fn=false_fn, strict=True, name=name)
F:Anaconda3envstflibsite-packagestensorflowpythonutildispatch.py:201 wrapper
    return target(*args, **kwargs)
F:Anaconda3envstflibsite-packagestensorflowpythonutildeprecation.py:507 new_func
    return func(*args, **kwargs)
F:Anaconda3envstflibsite-packagestensorflowpythonopscontrol_flow_ops.py:1180 cond
    return cond_v2.cond_v2(pred, true_fn, false_fn, name)
F:Anaconda3envstflibsite-packagestensorflowpythonopscond_v2.py:74 cond_v2
    pred = ops.convert_to_tensor(pred)
F:Anaconda3envstflibsite-packagestensorflowpythonframeworkops.py:1499 convert_to_tensor
    ret = conversion_func(value, dtype=dtype, name=name, as_ref=as_ref)
F:Anaconda3envstflibsite-packagestensorflowpythonframeworkconstant_op.py:338 _constant_tensor_conversion_function
    return constant(v, dtype=dtype, name=name)
F:Anaconda3envstflibsite-packagestensorflowpythonframeworkconstant_op.py:264 constant
    allow_broadcast=True)
F:Anaconda3envstflibsite-packagestensorflowpythonframeworkconstant_op.py:282 _constant_impl
    allow_broadcast=allow_broadcast))
F:Anaconda3envstflibsite-packagestensorflowpythonframeworktensor_util.py:444 make_tensor_proto
    raise ValueError("None values not supported.")

ValueError: None values not supported.
  

Я тоже пробовал training=False , но это тоже не работает.

Кажется, что Sequential() это хорошо работает с моим пользовательским слоем, но как я могу использовать его в своем формате