我正在尝试使用 MNIST 卷积神经网络训练结构，但我想进行计数，而不是进行分类，这意味着我的输出应该是标量而不是 softmax。我很难从sparse_softmax_cross_entropy移动到mean_squared_error作为我的损失函数。我得到的具体错误是：
值错误：形状 (100， 1( 和 (100，( 不兼容
但我也假设我的predictions函数设置不正确，因为它使用 argmax 和 softmax。我已经粘贴了下面的所有代码，但我认为问题出在损失函数和预测函数上。
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import numpy as np
import tensorflow as tf
tf.logging.set_verbosity(tf.logging.INFO)
def cnn_model_fn(features, labels, mode):
"""Model function for CNN."""
# Input Layer
# Reshape X to 4-D tensor: [batch_size, width, height, channels]
# Rack images are 116x116 pixels, and have one color channel
input_layer = tf.reshape(features["x"], [-1, 116, 116, 1])
# Convolutional Layer #1
# Computes 32 features using a 5x5 filter with ReLU activation.
# Padding is added to preserve width and height.
# Input Tensor Shape: [batch_size, 116, 116, 1]
# Output Tensor Shape: [batch_size, 116, 116, 32]
conv1 = tf.layers.conv2d(
inputs=input_layer,
filters=32,
kernel_size=[5, 5],
padding="same",
activation=tf.nn.relu)
# Pooling Layer #1
# First max pooling layer with a 2x2 filter and stride of 2
# Input Tensor Shape: [batch_size, 116, 116, 32]
# Output Tensor Shape: [batch_size, 58, 58, 32]
pool1 = tf.layers.max_pooling2d(inputs=conv1, pool_size=[2, 2], strides=2)
# Convolutional Layer #2
# Computes 64 features using a 5x5 filter.
# Padding is added to preserve width and height.
# Input Tensor Shape: [batch_size, 58, 58, 32]
# Output Tensor Shape: [batch_size, 58, 58, 64]
conv2 = tf.layers.conv2d(
inputs=pool1,
filters=64,
kernel_size=[5, 5],
padding="same",
activation=tf.nn.relu)
# Pooling Layer #2
# Second max pooling layer with a 2x2 filter and stride of 2
# Input Tensor Shape: [batch_size, 58, 58, 64]
# Output Tensor Shape: [batch_size, 29, 29, 64]
pool2 = tf.layers.max_pooling2d(inputs=conv2, pool_size=[2, 2], strides=2)
# Flatten tensor into a batch of vectors
# Input Tensor Shape: [batch_size, 29, 29, 256]
# Output Tensor Shape: [batch_size, 29 * 29 * 256]
pool2_flat = tf.reshape(pool2, [-1, 29 * 29 * 64])
# Dense Layer
# Densely connected layer with 1024 neurons
# Input Tensor Shape: [batch_size, 29 * 29 * 256]
# Output Tensor Shape: [batch_size, 1024]
dense = tf.layers.dense(inputs=pool2_flat, units=1024, activation=tf.nn.relu)
# Add dropout operation; 0.6 probability that element will be kept
dropout = tf.layers.dropout(
inputs=dense, rate=0.4, training=mode == tf.estimator.ModeKeys.TRAIN)
# Logits layer
# Input Tensor Shape: [batch_size, 1024]
# Output Tensor Shape: [batch_size, 5]
logits = tf.layers.dense(inputs=dropout, units=1)
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.mean_squared_error(labels=labels, predictions=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)
def main(unused_argv):
# Load training and eval data
train_data = np.load('Train_data.npy').astype(dtype=np.float32) # Returns np.array
train_labels = np.load('Train_labels.npy').astype(dtype=np.int32)
eval_data = np.load('Eval_data.npy').astype(dtype=np.float32)
eval_labels = np.load('Eval_labels.npy').astype(dtype=np.int32)
# Create the Estimator
mnist_classifier = tf.estimator.Estimator(
model_fn=cnn_model_fn)
# Set up logging for predictions
# Log the values in the "Softmax" tensor with label "probabilities"
tensors_to_log = {"probabilities": "softmax_tensor"}
logging_hook = tf.train.LoggingTensorHook(
tensors=tensors_to_log, every_n_iter=50)
# Train the model
train_input_fn = tf.estimator.inputs.numpy_input_fn(
x={"x": train_data},
y=train_labels,
batch_size=100,
num_epochs=None,
shuffle=True)
mnist_classifier.train(
input_fn=train_input_fn,
steps=5000,
hooks=[logging_hook])
# Evaluate the model and print results
eval_input_fn = tf.estimator.inputs.numpy_input_fn(
x={"x": eval_data},
y=eval_labels,
num_epochs=1,
shuffle=False)
eval_results = mnist_classifier.evaluate(input_fn=eval_input_fn)
print(eval_results)
# Evaluate single image
#single_image_predict = tf.estimator.inputs.numpy_input_fn(
#x = {"x":
if __name__ == "__main__":
tf.app.run()
完全错误：
File "C:/Users/smith25/Documents/Proof of Clean/Dish Machine Camera/Glasses MSE with OneHot.py", line 162, in <module>
tf.app.run()
File "C:Userssmith25AppDataLocalProgramsPythonPython36libsite-packagestensorflowpythonplatformapp.py", line 126, in run
_sys.exit(main(argv))
File "C:/Users/smith25/Documents/Proof of Clean/Dish Machine Camera/Glasses MSE with OneHot.py", line 145, in main
hooks=[logging_hook])
File "C:Userssmith25AppDataLocalProgramsPythonPython36libsite-packagestensorflowpythonestimatorestimator.py", line 363, in train
loss = self._train_model(input_fn, hooks, saving_listeners)
File "C:Userssmith25AppDataLocalProgramsPythonPython36libsite-packagestensorflowpythonestimatorestimator.py", line 843, in _train_model
return self._train_model_default(input_fn, hooks, saving_listeners)
File "C:Userssmith25AppDataLocalProgramsPythonPython36libsite-packagestensorflowpythonestimatorestimator.py", line 856, in _train_model_default
features, labels, model_fn_lib.ModeKeys.TRAIN, self.config)
File "C:Userssmith25AppDataLocalProgramsPythonPython36libsite-packagestensorflowpythonestimatorestimator.py", line 831, in _call_model_fn
model_fn_results = self._model_fn(features=features, **kwargs)
File "C:/Users/smith25/Documents/Proof of Clean/Dish Machine Camera/Glasses MSE with OneHot.py", line 100, in cnn_model_fn
loss = tf.losses.mean_squared_error(labels=labels, predictions=logits)
File "C:Userssmith25AppDataLocalProgramsPythonPython36libsite-packagestensorflowpythonopslosseslosses_impl.py", line 629, in mean_squared_error
predictions.get_shape().assert_is_compatible_with(labels.get_shape())
File "C:Userssmith25AppDataLocalProgramsPythonPython36libsite-packagestensorflowpythonframeworktensor_shape.py", line 844, in assert_is_compatible_with
raise ValueError("Shapes %s and %s are incompatible" % (self, other))
ValueError: Shapes (100, 1) and (100,) are incompatible
问题出在
loss = tf.losses.mean_squared_error(labels=labels, predictions=logits)
labels是形状(100，((即它只有1维，它是一个向量(，而logits是形状(100,1(的张量(它是一个二维矩阵，其中一个维度恰好是1(。看看它曾经是如何形状[batch_size, 5]的，现在它与 1 而不是 5 相同(。对于mean_squared_error的两个参数，你需要相同的形状，所以你需要做：
logits = tf.squeeze(tf.layers.dense(inputs=dropout, units=1)) # go from [batch_size, 1] to [batch_size]
此外，如果你现在正在做回归而不是分类(即试图预测一个数字，而不是一个类(，那么logits包含这个数字，"classes": tf.argmax(input=logits, axis=1)和"probabilities": tf.nn.softmax(logits, name="softmax_tensor")不再有任何意义，你可以删除它：logits 中只有 1 维(批处理维(。