我是tensorflow的初学者,我想使用自己的数据(40x40大小的图像)调整MNIST教程https://www.tensorflow.org/tutorials/layers。这是我的模型函数:
def cnn_model_fn(features, labels, mode):
# Input Layer
input_layer = tf.reshape(features, [-1, 40, 40, 1])
# Convolutional Layer #1
conv1 = tf.layers.conv2d(
inputs=input_layer,
filters=32,
kernel_size=[5, 5],
# To specify that the output tensor should have the same width and height values as the input tensor
# value can be "same" ou "valid"
padding="same",
activation=tf.nn.relu)
# Pooling Layer #1
pool1 = tf.layers.max_pooling2d(inputs=conv1, pool_size=[2, 2], strides=2)
# Convolutional Layer #2 and Pooling Layer #2
conv2 = tf.layers.conv2d(
inputs=pool1,
filters=64,
kernel_size=[5, 5],
padding="same",
activation=tf.nn.relu)
pool2 = tf.layers.max_pooling2d(inputs=conv2, pool_size=[2, 2], strides=2)
# Dense Layer
pool2_flat = tf.reshape(pool2, [-1, 10 * 10 * 64])
dense = tf.layers.dense(inputs=pool2_flat, units=1024, activation=tf.nn.relu)
dropout = tf.layers.dropout(
inputs=dense, rate=0.4, training=mode == tf.estimator.ModeKeys.TRAIN)
# Logits Layer
logits = tf.layers.dense(inputs=dropout, units=2)
predictions = {
# Generate predictions (for PREDICT and EVAL mode)
"classes": tf.argmax(input=logits, axis=1),
# Add `softmax_tensor` to the graph. It is used for PREDICT and by the
# `logging_hook`.
"probabilities": tf.nn.softmax(logits, name="softmax_tensor")
}
if mode == tf.estimator.ModeKeys.PREDICT:
return tf.estimator.EstimatorSpec(mode=mode, predictions=predictions)
# Calculate Loss (for both TRAIN and EVAL modes)
loss = tf.losses.sparse_softmax_cross_entropy(labels=labels, logits=logits)
# Configure the Training Op (for TRAIN mode)
if mode == tf.estimator.ModeKeys.TRAIN:
optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.001)
train_op = optimizer.minimize(
loss=loss,
global_step=tf.train.get_global_step())
return tf.estimator.EstimatorSpec(mode=mode, loss=loss, train_op=train_op)
# Add evaluation metrics (for EVAL mode)
eval_metric_ops = {
"accuracy": tf.metrics.accuracy(
labels=labels, predictions=predictions["classes"])}
return tf.estimator.EstimatorSpec(
mode=mode, loss=loss, eval_metric_ops=eval_metric_ops)
我在标签和逻辑回归之间遇到了一个形状大小错误:
InvalidArgumentError(有关详细信息,请参见上面的追溯):logits和labels必须具有相同的第一维,得到的logits形状为[3,2],标签形状为[1]
filenames_array是一个包含16个字符串的数组
["file1.png", "file2.png", "file3.png", ...]
并且 labels_array 是一个由16个整数组成的数组
[0,0,1,1,0,1,0,0,0,...]
主要功能是:
# Create the Estimator
mnist_classifier = tf.estimator.Estimator(model_fn=cnn_model_fn, model_dir="/tmp/test_convnet_model")
# Train the model
cust_train_input_fn = lambda: train_input_fn_custom(
filenames_array=filenames, labels_array=labels, batch_size=1)
mnist_classifier.train(
input_fn=cust_train_input_fn,
steps=20000,
hooks=[logging_hook])
我尝试重新调整logits的形状,但没有成功:
logits = tf.reshape(logits,[1, 2])
我需要你的帮助,谢谢
编辑
经过更多时间的搜索,在我的模型函数的第一行中:
input_layer = tf.reshape(features, [-1, 40, 40, 1])
“-1”代表批量大小的尺寸将动态计算,这里的值为“3”。这个“3”也出现在我的错误信息中:logits and labels must have the same first dimension, got logits shape [3,2] and labels shape [1]
如果我将这个值强制改为“1”,那么就会出现新的错误:
Input to reshape is a tensor with 4800 values, but the requested shape has 1600
可能是我的特征出了问题?
编辑2:
完整代码在此处:https://gist.github.com/geoffreyp/cc8e97aab1bff4d39e10001118c6322e
编辑3:
我已更新Gist。
logits = tf.layers.dense(inputs=dropout, units=1)
https://gist.github.com/geoffreyp/cc8e97aab1bff4d39e10001118c6322e
但是我并不完全理解您关于批次大小的答案,为什么批次大小可以是3,而我选择了批次大小为1?
如果我选择batch_size = 3,我会收到此错误: logits和labels必须具有相同的第一维,得到的logits形状为[9,1],标签形状为[3]
我尝试重新整形标签:
labels = tf.reshape(labels, [3, 1])
我更新了特征和标签的结构:
filenames_train = [['blackcorner-data/1.png', 'blackcorner-data/2.png', 'blackcorner-data/3.png',
'blackcorner-data/4.png', 'blackcorner-data/n1.png'],
['blackcorner-data/n2.png',
'blackcorner-data/n3.png', 'blackcorner-data/n4.png',
'blackcorner-data/11.png', 'blackcorner-data/21.png'],
['blackcorner-data/31.png',
'blackcorner-data/41.png', 'blackcorner-data/n11.png', 'blackcorner-data/n21.png',
'blackcorner-data/n31.png']
]
labels = [[0, 0, 0, 0, 1], [1, 1, 1, 0, 0], [0, 0, 1, 1, 1]]
但是没有成功...