tensorflow - Python 中的错误：函数调用堆栈：train_function

Question

我正在 COLAB 工作环境中用 Python 语言运行一段应该训练手稿识别模型的代码片段。该代码加载了一个由 MNIST 人员撰写的文学图片​​数据库并进行练习。编码：

import numpy
from keras.datasets import mnist
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
from tensorflow.keras.layers import Dropout
from tensorflow.keras.layers import Flatten
from tensorflow.keras.layers import Conv2D
from tensorflow.keras.layers import MaxPooling2D
from tensorflow.keras import utils as np_utils
from tensorflow.keras import backend as K
K.set_image_data_format('channels_first')
from matplotlib import pyplot as plt

#load data from mnist
(xTrain,yTrain),(xTest,yTest)=mnist.load_data()
#reshape the images to be 28*28 pixels
xTrain=xTrain.reshape(xTrain.shape[0],1,28,28).astype('float32')
xTest=xTest.reshape(xTest.shape[0],1,28,28).astype('float32')
#normalize inputs from 0-255 to 0-1
xTrain=xTrain/255
xTest=xTest/255
#one hot encode outputs
yTrain=np_utils.to_categorical(yTrain)
yTest=np_utils.to_categorical(yTest)
num_classes=yTest.shape[1]

    
def baseline_model():
  #create model
  model=Sequential()
  model.add(Conv2D(32,(5,5), input_shape=(1,28,28),activation='relu'))
  model.add(MaxPooling2D(pool_size=(2,2)))
  model.add(Dropout(0.2))
  model.add(Flatten())
  model.add(Dense(128,activation='relu'))
  model.add(Dense(num_classes, activation='softmax'))
  #complie model
  model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
  return model


model=baseline_model()
model.fit(xTrain, yTrain, validation_data=(xTest, yTest), epochs=1, batch_size=200, verbose=2)

问题是代码在最后一行返回错误。错误：

    InvalidArgumentError                      Traceback (most recent call last)
<ipython-input-111-ec22b5dcc4e3> in <module>()
     63 
     64 model=baseline_model()
---> 65 model.fit(xTrain, yTrain, validation_data=(xTest, yTest), epochs=1, batch_size=200, verbose=2)

6 frames
/usr/local/lib/python3.7/dist-packages/tensorflow/python/eager/execute.py in quick_execute(op_name, num_outputs, inputs, attrs, ctx, name)
     58     ctx.ensure_initialized()
     59     tensors = pywrap_tfe.TFE_Py_Execute(ctx._handle, device_name, op_name,
---> 60                                         inputs, attrs, num_outputs)
     61   except core._NotOkStatusException as e:
     62     if name is not None:

InvalidArgumentError:  Default MaxPoolingOp only supports NHWC on device type CPU
     [[node sequential_36/max_pooling2d_21/MaxPool (defined at <ipython-input-111-ec22b5dcc4e3>:65) ]] [Op:__inference_train_function_10696]

Function call stack:
train_function

非常感谢帮助，谢谢

标签： tensorflowgoogle-colaboratory

Answer 1

InvalidArgumentError：默认 MaxPoolingOp 仅在设备类型 CPU 上支持 NHWC

正如错误清楚地表明的那样，它不能与 CPU 上的 NCHW（即 batch_size、channels、height、width）一起使用。

如果要使用 CPU，则应将格式更改为 NHWC (iebatch_size, height, width, channels)，然后输入应为 (28,28,1)。

还有一个建议永远不要混合来自 tensorflow 和 keras 的导入。

在 GPU 上，它支持 NCHW。您可以参考工作代码，如下所示

import numpy
from tensorflow.keras.datasets import mnist # import from TF
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
from tensorflow.keras.layers import Dropout
from tensorflow.keras.layers import Flatten
from tensorflow.keras.layers import Conv2D
from tensorflow.keras.layers import MaxPooling2D
from tensorflow.keras import utils as np_utils
from tensorflow.keras import backend as K
K.set_image_data_format('channels_first')
from matplotlib import pyplot as plt

#load data from mnist
(xTrain,yTrain),(xTest,yTest)=mnist.load_data()
#reshape the images to be 28*28 pixels
xTrain=xTrain.reshape(xTrain.shape[0],1,28,28).astype('float32')
xTest=xTest.reshape(xTest.shape[0],1,28,28).astype('float32')
#normalize inputs from 0-255 to 0-1
xTrain=xTrain/255
xTest=xTest/255
#one hot encode outputs
yTrain=np_utils.to_categorical(yTrain)
yTest=np_utils.to_categorical(yTest)
num_classes=yTest.shape[1]

    
def baseline_model():
  #create model
  model=Sequential()
  model.add(Conv2D(32,(5,5), input_shape=(1,28,28),activation='relu'))
  model.add(MaxPooling2D(pool_size=(2,2)))
  model.add(Dropout(0.2))
  model.add(Flatten())
  model.add(Dense(128,activation='relu'))
  model.add(Dense(num_classes, activation='softmax'))
  #complie model
  model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
  return model


model=baseline_model()
model.fit(xTrain, yTrain, validation_data=(xTest, yTest), epochs=1, batch_size=200, verbose=2)

输出：

Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz
11493376/11490434 [==============================] - 0s 0us/step
300/300 - 34s - loss: 0.2372 - accuracy: 0.9304 - val_loss: 0.0756 - val_accuracy: 0.9760