ValueError：无法为张量 u'InputData/X:0' 提供形状 (2, 4) 的值，其形状为 '(?, 2, 4

Question

我收到一个错误，ValueError：无法为形状为“(?, 2, 4, 104)”的张量 u'InputData/X:0' 提供形状 (2, 4) 的值。 我写了代码，

# coding: utf-8
import tensorflow as tf
import tflearn

from tflearn.layers.core import input_data,dropout,fully_connected
from tflearn.layers.conv import conv_2d, max_pool_2d
from tflearn.layers.normalization import local_response_normalization
from tflearn.layers.estimator import regression

import pandas as pd
import numpy as np
from sklearn import metrics

tf.reset_default_graph()
net = input_data(shape=[2, 4, 104])
net = conv_2d(net, 4, 16, activation='relu')
net = max_pool_2d(net, 1)
net = tflearn.activations.relu(net)
net = dropout(net, 0.5)
net = tflearn.fully_connected(net, 10, activation='softmax')
net = tflearn.regression(net, optimizer='adam', learning_rate=0.5, loss='categorical_crossentropy')

model = tflearn.DNN(net)

trainDataSet = [[0.25,0.25,1,1],[0,0,1,1],[0.25,0.25,1,1]]
trainLabel = [[0,1],[0,1],[1,0]]
model.fit(trainDataSet, trainLabel, n_epoch=100, batch_size=32, validation_set=0.1, show_metric=True)

回溯说

Traceback (most recent call last):
  File "cnn.py", line 16, in <module>
    model.fit(trainDataSet, trainLabel, n_epoch=100, batch_size=32, validation_set=0.1, show_metric=True)
  File "/Users/xxx/anaconda/xxx/lib/python2.7/site-packages/tflearn/models/dnn.py", line 216, in fit
    callbacks=callbacks)
  File "/Users/xxx/anaconda/xxx/lib/python2.7/site-packages/tflearn/helpers/trainer.py", line 339, in fit
    show_metric)
  File "/Users/xxx/anaconda/xxx/lib/python2.7/site-packages/tflearn/helpers/trainer.py", line 818, in _train
    feed_batch)
  File "/Users/xxx/anaconda/xxx/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 929, in run
    run_metadata_ptr)
  File "/Users/xxx/anaconda/xxx/lib/python2.7/site-packages/tensorflow/python/client/session.py", line 1128, in _run
    str(subfeed_t.get_shape())))
ValueError: Cannot feed value of shape (2, 4) for Tensor u'InputData/X:0', which has shape '(?, 2, 4, 104)'

我改写成

trainDataSet = np.array([[0.25,0.25,1,1],[0,0,1,1],[0.25,0.25,1,1]])
trainLabel = np.array([[0,1],[0,1],[1,0]])

但同样的错误发生了。我的代码有什么问题？我应该如何解决这个问题？

Answer 1

引用Tensorflow 文档：

tflearn.layers.conv.conv_2d 

输入：

4-D 张量 [batch, height, width, in_channels]。

来自其他 TensorFlow 文档：

tf.nn.conv2d

在给定4-D输入和过滤张量的情况下计算2-D卷积。

给定一个形状为 [batch, in_height, in_width, in_channels] 和形状为 [filter_height, filter_width, in_channels, out_channels]，这个操作执行 以下：

您的数据集、标签和输入形状未对齐，即彼此不匹配。

目前您的trainDataSet 的形状为 (3,4)：

import numpy as np
trainDataSet = np.array([[0.25,0.25,1,1],[0,0,1,1],[0.25,0.25,1,1]])
print(trainDataSet.shape)

输出：

(3, 4)

但是您将输入形状定义为：

net = input_data(shape=[2, 4, 104])

不明确你真正想要实现的目标，但如果你想看一个简单的工作示例，那么你的代码应该如下所示：

import tensorflow as tf
import tflearn

from tflearn.layers.core import input_data,dropout,fully_connected
from tflearn.layers.conv import conv_2d, max_pool_2d
from tflearn.layers.normalization import local_response_normalization
from tflearn.layers.estimator import regression

import pandas as pd
import numpy as np
from sklearn import metrics

tf.reset_default_graph()
net = input_data(shape=[3, 4, 1])
net = conv_2d(net, 4, 16, activation='relu')
net = max_pool_2d(net, 1)
net = tflearn.activations.relu(net)
net = dropout(net, 0.5)
net = tflearn.fully_connected(net, 2, activation='softmax')
net = tflearn.regression(net, optimizer='adam', learning_rate=0.5, loss='categorical_crossentropy')

model = tflearn.DNN(net)

trainDataSet = [
    [
        [[0.25], [0.25], [1], [1]],
        [[0], [0], [1], [1]],
        [[0.25], [0.25], [1], [1]]
    ],
    [
        [[0.25], [0.25], [1], [1]],
        [[0], [0], [1], [1]],
        [[0.25], [0.25], [1], [1]]
    ],
    [
        [[0.25], [0.25], [1], [1]],
        [[0], [0], [1], [1]],
        [[0.25], [0.25], [1], [1]]
    ]
]

trainLabel = [[0,1],[0,1],[1,0]]
model.fit(trainDataSet, trainLabel, n_epoch=100, batch_size=32, validation_set=0.1, show_metric=True)

输出：

---------------------------------
Run id: NHHJV7
Log directory: /tmp/tflearn_logs/
INFO:tensorflow:Summary name Accuracy/ (raw) is illegal; using Accuracy/__raw_ instead.
---------------------------------
Training samples: 2
Validation samples: 1
--
Training Step: 1  | time: 1.160s
| Adam | epoch: 001 | loss: 0.00000 - acc: 0.0000 | val_loss: 23.02585 - val_acc: 0.0000 -- iter: 2/2
--
Training Step: 2  | total loss: 0.62966 | time: 1.008s
| Adam | epoch: 002 | loss: 0.62966 - acc: 0.0000 | val_loss: 10.76885 - val_acc: 0.0000 -- iter: 2/2
.
.
.
Training Step: 99  | total loss: 0.00000 | time: 1.013s
| Adam | epoch: 099 | loss: 0.00000 - acc: 1.0000 | val_loss: 23.02585 - val_acc: 0.0000 -- iter: 2/2
--
Training Step: 100  | total loss: 0.00000 | time: 1.011s
| Adam | epoch: 100 | loss: 0.00000 - acc: 1.0000 | val_loss: 23.02585 - val_acc: 0.0000 -- iter: 2/2
--