UnimplementedError：图形执行错误：在TensorFlow上运行NN

Question

我一直在遇到这个错误，我不知道为什么，尤其是因为我完全关注某人的代码，并且运行时没有错误，这

img_shape = (128,128,3)

# load pretrained model
base_model = tf.keras.applications.VGG19(input_shape=img_shape, include_top=False, weights='imagenet')

# freezing the model
base_model.trainable = False

#define the custom head for network
global_average_layer = tf.keras.layers.GlobalAveragePooling2D()(base_model.output)

# output / prediction layer
prediction_layer = tf.keras.layers.Dense(units=1, activation='sigmoid')(global_average_layer)

model = tf.keras.models.Model(inputs=base_model.input, outputs=prediction_layer)

# compile the model
opt = tf.keras.optimizers.RMSprop(learning_rate=0.0001)
model.compile(optimizer=opt, loss='binary_crossentropy', metrics=['accuracy'])

# create data generators
# import library
from tensorflow.keras.preprocessing.image import ImageDataGenerator

# define objects
data_gen_train = ImageDataGenerator(rescale=1/255.0)
data_gen_test = ImageDataGenerator(rescale=1/255.0)

# define variables
train_generator = data_gen_train.flow_from_directory(directory=training_dir, target_size=(128,128), batch_size=128, class_mode='binary')
test_generator = data_gen_test.flow_from_directory(directory=test_dir, target_size=(128,128), batch_size=128, class_mode='binary')

model.fit_generator(generator=train_generator, epochs=5, validation_data=test_generator)

是我遇到的错误

/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:1: UserWarning: `Model.fit_generator` is deprecated and will be removed in a future version. Please use `Model.fit`, which supports generators.
  """Entry point for launching an IPython kernel.
Epoch 1/5
---------------------------------------------------------------------------
UnimplementedError                        Traceback (most recent call last)
<ipython-input-46-18b18ca5977c> in <module>()
----> 1 model.fit_generator(generator=train_generator, epochs=5, validation_data=test_generator)

2 frames
/usr/local/lib/python3.7/dist-packages/keras/engine/training.py in fit_generator(self, generator, steps_per_epoch, epochs, verbose, callbacks, validation_data, validation_steps, validation_freq, class_weight, max_queue_size, workers, use_multiprocessing, shuffle, initial_epoch)
   2221         use_multiprocessing=use_multiprocessing,
   2222         shuffle=shuffle,
-> 2223         initial_epoch=initial_epoch)
   2224 
   2225   @doc_controls.do_not_generate_docs

/usr/local/lib/python3.7/dist-packages/keras/utils/traceback_utils.py in error_handler(*args, **kwargs)
     65     except Exception as e:  # pylint: disable=broad-except
     66       filtered_tb = _process_traceback_frames(e.__traceback__)
---> 67       raise e.with_traceback(filtered_tb) from None
     68     finally:
     69       del filtered_tb
/usr/local/lib/python3.7/dist-packages/tensorflow/python/eager/execute.py in quick_execute(op_name, num_outputs, inputs, attrs, ctx, name)
     53     ctx.ensure_initialized()
     54     tensors = pywrap_tfe.TFE_Py_Execute(ctx._handle, device_name, op_name,
---> 55                                         inputs, attrs, num_outputs)
     56   except core._NotOkStatusException as e:
     57     if name is not None:

UnimplementedError: Graph execution error:

Detected at node 'model/block1_conv1/Conv2D' defined at (most recent call last):
    File "/usr/lib/python3.7/runpy.py", line 193, in _run_module_as_main
      "__main__", mod_spec)

，我无法在结尾处遇到完全错误， 但 拜托，有人可以告诉我怎么了吗

Answer 1

很容易使用train_generator旧，您可以使用它。

它是关于您创建矩阵初始操作的内存，该操作将减少工作负载并在数据尺寸上表现良好。

[示例]：

import os
from os.path import exists

import tensorflow as tf

"""""""""""""""""""""""""""""""""""""""""""""""""""""""""
: Variables
"""""""""""""""""""""""""""""""""""""""""""""""""""""""""
img_shape = (128,128,3)
BATCH_SIZE = 1
IMG_SIZE = (128, 128)

database_buffer = "F:\\models\\buffer\\" + os.path.basename(__file__).split('.')[0] + "\\TF_DataSets_01.h5"
database_buffer_dir = os.path.dirname(database_buffer)

if not exists(database_buffer_dir) : 
    os.mkdir(database_buffer_dir)
    print("Create directory: " + database_buffer_dir)

"""""""""""""""""""""""""""""""""""""""""""""""""""""""""
: DataSets
"""""""""""""""""""""""""""""""""""""""""""""""""""""""""
PATH = 'F:\\datasets\\downloads\\cats_name'
train_dir = os.path.join(PATH, 'train')
validation_dir = os.path.join(PATH, 'validation')

train_dataset = tf.keras.utils.image_dataset_from_directory(train_dir,
                                                                        shuffle=True,
                                                                        batch_size=BATCH_SIZE,
                                                                        image_size=IMG_SIZE)

"""""""""""""""""""""""""""""""""""""""""""""""""""""""""
: Model Initialize
"""""""""""""""""""""""""""""""""""""""""""""""""""""""""
# load pretrained model
base_model = tf.keras.applications.VGG19(input_shape=img_shape, include_top=False, weights='imagenet')

# freezing the model
base_model.trainable = False

#define the custom head for network
global_average_layer = tf.keras.layers.GlobalAveragePooling2D()(base_model.output)

# output / prediction layer
prediction_layer = tf.keras.layers.Dense(units=1, activation='sigmoid')(global_average_layer)

model = tf.keras.models.Model(inputs=base_model.input, outputs=prediction_layer)
model.summary()

# compile the model
opt = tf.keras.optimizers.RMSprop(learning_rate=0.0001)
model.compile(optimizer=opt, loss='binary_crossentropy', metrics=['accuracy'])

"""""""""""""""""""""""""""""""""""""""""""""""""""""""""
: Training
"""""""""""""""""""""""""""""""""""""""""""""""""""""""""
history = model.fit( train_dataset, batch_size=100, epochs=50 )

[output]：

sample << /a>

Answer 2

如果您使用的是Colab，请尝试将GPU切换到TPU。这可能会有所帮助

Answer 3

使用jit_compile = false，这对我有用：

my_model.compile(optimizer="Adam",...., jit_compile=False)

Answer 4

解决此错误的解决方案是检查图像形状，并且您还为神经网络添加了形状。两者都必须相同。

这解决了我的问题。