我在尝试使用tensorflow实现一个dcgan时遇到了这个错误:
ValueError: Shapes must be equal rank, but are 2 and 1From merging shape 1 with other shapes. for 'generator/Reshape/packed' (op: 'Pack') with input shapes: [?,2048], [100,2048], [2048].
据我所知,这表明我的张量形状不同,但我看不出需要更改什么来修复这个错误。我认为错误可能出现在以下方法之间:
首先,我在一个方法中创建了一个占位符,使用:
self.z = tf.placeholder(tf.float32, [None,self.z_dimension], name='z')self.z_sum = tf.histogram_summary("z", self.z)self.G = self.generator(self.z)
然后最后一个语句调用了生成器方法,该方法通过reshape更改张量:
self.z_ = linear(z,self.gen_dimension * 8 * sample_H16 * sample_W16, 'gen_h0_lin', with_w=True) self.h0 = tf.reshape(self.z_,[-1, sample_H16, sample_W16,self.gen_dimension * 8]) h0 = tf.nn.relu(self.gen_batchnorm1(self.h0))
如果有帮助,这里是我的线性方法:
def linear(input_, output_size, scope=None, stddev=0.02, bias_start=0.0, with_w=False):shape = input_.get_shape().as_list()with tf.variable_scope(scope or "Linear"): matrix = tf.get_variable("Matrix", [shape[1], output_size], tf.float32,tf.random_normal_initializer(stddev=stddev)) bias = tf.get_variable("bias", [output_size],initializer=tf.constant_initializer(bias_start)) if with_w: return tf.matmul(input_, matrix) + bias, matrix, bias else: return tf.matmul(input_, matrix) + bias
编辑:
我还使用了这些占位符:
self.inputs = tf.placeholder(tf.float32, shape=[self.batch_size] + image_dimension, name='real_images') self.gen_inputs = tf.placeholder(tf.float32, shape=[self.sample_size] + image_dimension, name='sample_inputs') inputs = self.inputs sample_inputs = self.gen_inputs
回答:
linear(z, self.gen_dimension * 8 * sample_H16 * sample_W16, 'gen_h0_lin', with_w=True)
将返回元组 (tf.matmul(input_, matrix) + bias, matrix, bias)
。
因此,self.z_
被赋值为元组,而不是单个张量。
只需将 linear(z, self.gen_dimension * 8 * sample_H16 * sample_W16, 'gen_h0_lin', with_w=True)
更改为 linear(z, self.gen_dimension * 8 * sample_H16 * sample_W16, 'gen_h0_lin', with_w=False)
即可。