我对张量流、keras 和深度学习场景很陌生。
我尝试制作一个简单的神经网络来识别 fer2013 数据库中的情绪，但在训练第一个 epoch 后不断遇到此错误
File "main.py", line 92, in <module>
train(image_data, label_data, conv_arch, dense, dropout, epochs, batch_size, validation_split, patience)
File "main.py", line 38, in train
hist = model.fit(image_data, label_data, epochs=epochs, batch_size=batch_size, validation_split=validation_split, callbacks=callbacks, shuffle=True, verbose=1)
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/keras/models.py", line 965, in fit
validation_steps=validation_steps)
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/keras/engine/training.py", line 1669, in fit
validation_steps=validation_steps)
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/keras/engine/training.py", line 1226, in _fit_loop
callbacks.on_epoch_end(epoch, epoch_logs)
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/keras/callbacks.py", line 76, in on_epoch_end
callback.on_epoch_end(epoch, logs)
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/keras/callbacks.py", line 786, in on_epoch_end
result = self.sess.run([self.merged], feed_dict=feed_dict)
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 767, in run
run_metadata_ptr)
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 965, in _run
feed_dict_string, options, run_metadata)
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1015, in _do_run
target_list, options, run_metadata)
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1035, in _do_call
raise type(e)(node_def, op, message)
tensorflow.python.framework.errors_impl.InvalidArgumentError: You must feed a value for placeholder tensor 'conv2d_1_input' with dtype float
[[Node: conv2d_1_input = Placeholder[dtype=DT_FLOAT, shape=[], _device="/job:localhost/replica:0/task:0/cpu:0"]()]]
Caused by op u'conv2d_1_input', defined at:
File "main.py", line 92, in <module>
train(image_data, label_data, conv_arch, dense, dropout, epochs, batch_size, validation_split, patience)
File "main.py", line 25, in train
model = create_model(image_data, label_data, conv_arch=conv_arch, dense=dense, dropout=dropout)
File "/Users/rifkibramantyo/tensorflow/variance/src/deepCNN.py", line 22, in create_model
model.add(Conv2D(conv_arch[0][0], kernel_size=(3, 3), strides=3, padding='same', activation='relu', data_format='channels_first', input_shape=(1, image_data_train.shape[2], image_data_train.shape[3])))
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/keras/models.py", line 463, in add
name=layer.name + '_input')
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/keras/engine/topology.py", line 1455, in Input
input_tensor=tensor)
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/keras/legacy/interfaces.py", line 91, in wrapper
return func(*args, **kwargs)
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/keras/engine/topology.py", line 1364, in __init__
name=self.name)
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/keras/backend/tensorflow_backend.py", line 504, in placeholder
x = tf.placeholder(dtype, shape=shape, name=name)
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/ops/array_ops.py", line 1520, in placeholder
name=name)
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/ops/gen_array_ops.py", line 2149, in _placeholder
name=name)
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py", line 763, in apply_op
op_def=op_def)
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2395, in create_op
original_op=self._default_original_op, op_def=op_def)
File "/Users/rifkibramantyo/anaconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1264, in __init__
self._traceback = _extract_stack()
InvalidArgumentError (see above for traceback): You must feed a value for placeholder tensor 'conv2d_1_input' with dtype float
[[Node: conv2d_1_input = Placeholder[dtype=DT_FLOAT, shape=[], _device="/job:localhost/replica:0/task:0/cpu:0"]()]]
我试图摆弄代码，当我删除回调数组时，错误消失了main.py？我的回调格式是否错误？
这是源代码
main.py
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
from datetime import datetime
import numpy as np
from keras.callbacks import EarlyStopping, TensorBoard
from keras.utils import plot_model
from keras import backend as K
from deepCNN import *
from log import *
K.set_learning_phase(1)
def load_test_data():
image_data_fname = '../data/data_images_Training.npy'
label_data_fname = '../data/data_labels_Training.npy'
image_data = np.load(image_data_fname)
label_data = np.load(label_data_fname)
return image_data, label_data
def train(image_data, label_data, conv_arch, dense, dropout, epochs, batch_size, validation_split, patience):
model = create_model(image_data, label_data, conv_arch=conv_arch, dense=dense, dropout=dropout)
plot_model(model, to_file="graph.png")
# define callbacks
callbacks = []
if patience != 0:
early_stopping = EarlyStopping(monitor='val_loss', patience=patience, verbose=1)
tensor_board = TensorBoard(log_dir='../logs', histogram_freq=1, write_graph=True, write_grads=False, write_images=False, embeddings_freq=0, embeddings_layer_names=None, embeddings_metadata=None)
callbacks.append(early_stopping)
callbacks.append(tensor_board)
start_time = datetime.now()
print('Training....\n')
hist = model.fit(image_data, label_data, epochs=epochs, batch_size=batch_size, validation_split=validation_split, callbacks=callbacks, shuffle=True, verbose=1)
end_time = datetime.now()
# model result:
train_val_accuracy = hist.history
train_acc = train_val_accuracy['acc']
val_acc = train_val_accuracy['val_acc']
print(' Done!')
print(' Train acc: ', train_acc[-1])
print('Validation acc: ', val_acc[-1])
print(' Overfit ratio: ', val_acc[-1] / train_acc[-1])
my_log(model, start_time, end_time,
batch_size=batch_size,
epochs=epochs,
conv_arch=conv_arch,
dense=dense,
dropout=dropout,
image_data_shape=image_data.shape,
train_acc=train_acc,
val_acc=val_acc,
dirpath=result_location)
if __name__ == '__main__':
image_data, label_data = load_test_data()
result_location = '../data/results/'
arr_conv_arch = [[(32, 1), (64, 0), (128, 0)],
[(32, 0), (64, 1), (128, 0)],
[(32, 0), (64, 0), (128, 1)],
[(32, 2), (64, 0), (128, 0)],
[(32, 3), (64, 0), (128, 0)],
[(32, 1), (64, 1), (128, 1)],
[(32, 2), (64, 2), (128, 1)],
[(32, 3), (64, 2), (128, 1)],
[(32, 3), (64, 3), (128, 3)]]
arr_dense = [[64, 2], (256,2),(256,4),(512,2),(512,4)]
arr_dropouts = [0.2, 0.3, 0.4, 0.5]
arr_epochs = [1, 10, 40]
arr_batch_size = [50, 256]
validation_split = 0.2
patience = 5
#train(image_data, label_data, arr_conv_arch[8], arr_dense[0], arr_dropouts[0], 1, arr_batch_size[0], validation_split, patience)
#train(image_data, label_data, arr_conv_arch[1], arr_dense[1], arr_dropouts[0], arr_epochs[0], arr_batch_size[0], validation_split, patience)
for conv_arch in arr_conv_arch:
for dense in arr_dense:
for dropout in arr_dropouts:
for epochs in arr_epochs:
for batch_size in arr_batch_size:
print_architecture(image_data.shape, label_data.shape, batch_size, dropout, epochs, conv_arch, dense)
train(image_data, label_data, conv_arch, dense, dropout, epochs, batch_size, validation_split, patience)
深度CNN.py
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import os
# disable some warnings
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
from keras.layers import Conv2D, MaxPooling2D
from keras.layers import Dropout, Flatten, Dense
from keras.models import Sequential
from keras.layers.core import K
K.set_learning_phase(1)
def create_model(image_data_train, label_data_train, conv_arch=[(32, 1)], dense=[32, 1], dropout=0.5):
image_data_train = image_data_train.astype('float32')
# Define model:
model = Sequential()
model.add(Conv2D(conv_arch[0][0], kernel_size=(3, 3), strides=3, padding='same', activation='relu', data_format='channels_first', input_shape=(1, image_data_train.shape[2], image_data_train.shape[3])))
if (conv_arch[0][1] - 1) != 0:
for i in range(conv_arch[0][1] - 1):
model.add(Conv2D(conv_arch[0][0], kernel_size=(3, 3), strides=1, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
if conv_arch[1][1] != 0:
for i in range(conv_arch[1][1]):
model.add(Conv2D(conv_arch[1][0], kernel_size=(3, 3), strides=1, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
if conv_arch[2][1] != 0:
for i in range(conv_arch[2][1]):
model.add(Conv2D(conv_arch[2][0], kernel_size=(3, 3), strides=1, padding='same', activation='relu'))
model.add(MaxPooling2D(pool_size=2))
# Print data shape through network for debugging
for z in range(len(model.layers)):
print("{} : {} --> {}".format(z, model.layers[z].input_shape, model.layers[z].output_shape))
model.add(Flatten()) # this converts 3D feature maps to 1D feature vectors
if dense[1] != 0:
for i in range(dense[1]):
model.add(Dense(dense[0], activation='relu'))
if dropout:
model.add(Dropout(dropout))
prediction = model.add(Dense(label_data_train.shape[1], activation='softmax'))
# optimizer:
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
return model
log.py
import os
import pandas as pd
def print_architecture(image_data_shape, label_data_shape, batch_size, dropout, epochs, conv_arch, dense):
print(' image data shape: ', image_data_shape) # (n_sample, 1, 48, 48)
print(' label data shape: ', label_data_shape) # (n_sample, n_categories)
print(' img size: ', image_data_shape[2], image_data_shape[3])
print(' batch size: ', batch_size)
print(' epochs: ', epochs)
print(' dropout: ', dropout)
print('conv architect: ', conv_arch)
print('neural network: ', dense)
def time_difference(start, end):
return '{}_min_{}_sec'.format(end.time().minute - start.time().minute,
end.time().second - start.time().second)
def save_config(formatted_val_acc, start_time, config, dirpath='../data/results/'):
with open(dirpath + formatted_val_acc + '-' + 'config_log.txt', 'a') as f:
f.write(start_time.isoformat() + '\n')
f.write(str(config) + '\n')
def save_model(formatted_val_acc, start_time, json_string, dirpath='../data/results/'):
with open(dirpath + formatted_val_acc + '-' + start_time.isoformat() + '_model.txt', 'w') as f:
f.write(json_string)
def save_result(start_time, end_time, train_acc, val_acc, conv_arch, batch_size, epochs, dense, dropout, dirpath='../data/results/'):
with open(dirpath + 'result_log.txt', 'a') as f:
f.write(start_time.isoformat() + ' -> ' + end_time.isoformat() + '\n')
f.write(' batch size: ' + str(batch_size) +
'\n epochs: ' + str(epochs) +
'\n dropout: ' + str(dropout) +
'\n conv: ' + str(conv_arch) +
'\n dense: ' + str(dense) +
'\n - Train acc: ' + str(train_acc[-1]) +
'\n - Val acc: ' + str(val_acc[-1]) +
'\n - Ratio: ' + str(val_acc[-1] / train_acc[-1]) + '\n\n')
# def append_to_result_log(text):
# with open('../data/results/result_log.txt', 'a') as f:
# f.write('\n' + text + '\n')
def my_log(model, start_time, end_time, batch_size, epochs, conv_arch, dense, dropout, image_data_shape, train_acc, val_acc, dirpath):
if (val_acc[-1] > 0.5):
formatted_val_acc = "{:.4f}".format(val_acc[-1])
model.save_weights('../data/weights/{}-{}'.format(formatted_val_acc, start_time))
save_model(formatted_val_acc, start_time, model.to_json(), dirpath)
save_config(formatted_val_acc, start_time, model.get_config(), dirpath)
save_result(start_time, end_time, train_acc, val_acc, conv_arch, batch_size, epochs, dense, dropout, dirpath)
save_as_csv(start_time, end_time, train_acc, val_acc, conv_arch, batch_size, epochs, dense, dropout, image_data_shape, dirpath)
def save_as_csv(start_time, end_time, train_acc, val_acc, conv_arch, batch_size, epochs, dense, dropout, image_data_shape, dirpath):
data = {
'start_time': start_time,
'end_time': end_time,
'duration': time_difference(start_time, end_time),
'batch_size': batch_size,
'epochs': epochs,
'conv_arch': conv_arch,
'dense': dense,
'dropout': dropout,
'image_width': image_data_shape[2],
'image_height': image_data_shape[3],
'train_acc': train_acc,
'val_acc': val_acc
}
df = pd.DataFrame(dict([(k, pd.Series(v)) for k, v in data.iteritems()]))
filepath = dirpath + 'results.csv'
if not os.path.isfile(filepath):
df.to_csv(filepath, index=False, sep='#', escapechar='^', header='column_names')
else: # else it exists so append without writing the header
df.to_csv(filepath, index=False, sep='#', escapechar='^', mode='a', header=False)
