将列表输入 TensorFlow 中的 feed_dict 时发出问题 [英] Issue feeding a list into feed_dict in TensorFlow

问题描述

我正在尝试将列表传递到 feed_dict，但是我在这样做时遇到了问题.说我有:

I'm trying to pass a list into feed_dict, however I'm having trouble doing so. Say I have:

inputs = 10 * [tf.placeholder(tf.float32, shape=(batch_size, input_size))]

其中输入被输入到我想要计算的某个函数 outputs 中.因此，为了在 tensorflow 中运行它，我创建了一个会话并运行以下命令:

where inputs is fed into some function outputs that I want to compute. So to run this in tensorflow, I created a session and ran the following:

sess.run(outputs, feed_dict = {inputs: data}) 
#data is my list of inputs, which is also of length 10

但我得到一个错误，TypeError: unhashable type: 'list'.但是，我可以像这样按元素传递数据:

but I get an error, TypeError: unhashable type: 'list'. However, I'm able to pass the data element-wise like so:

sess.run(outputs, feed_dict = {inputs[0]: data[0], ..., inputs[9]: data[9]}) 

所以我想知道是否有办法解决这个问题.我还尝试构建一个字典(使用 for 循环)，但是这会生成一个包含单个元素的字典，它们的键是:tensorflow.python.framework.ops.Tensor at 0x107594a10

So I'm wondering if there's a way I can solve this issue. I've also tried to construct a dictionary(using a for loop), however this results in a dictionary with a single element, where they key is: tensorflow.python.framework.ops.Tensor at 0x107594a10

推荐答案

这里有两个问题导致了问题:

There are two issues that are causing problems here:

第一个问题是 Session.run() 调用只接受少量类型作为 feed_dict 的键.特别是，张量列表不支持作为键，因此您必须将每个张量作为单独的键.^*一种方便的方法这样做是使用字典理解:

The first issue is that the Session.run() call only accepts a small number of types as the keys of the feed_dict. In particular, lists of tensors are not supported as keys, so you have to put each tensor as a separate key.^* One convenient way to do this is using a dictionary comprehension:

inputs = [tf.placeholder(...), ...]
data = [np.array(...), ...]
sess.run(y, feed_dict={i: d for i, d in zip(inputs, data)})

第二个问题是 Python 中的 10 * [tf.placeholder(...)] 语法创建了一个包含十个元素的列表，其中每个元素相同张量对象(即具有相同的 name 属性，相同的 id 属性，并且如果比较列表中的两个元素，则引用相同使用 inputs[i] 是输入 [j]).这解释了为什么当您尝试使用列表元素作为键创建字典时，您最终会得到一个包含单个元素的字典 - 因为所有列表元素都是相同的.

The second issue is that the 10 * [tf.placeholder(...)] syntax in Python creates a list with ten elements, where each element is the same tensor object (i.e. has the same name property, the same id property, and is reference-identical if you compare two elements from the list using inputs[i] is inputs[j]). This explains why, when you tried to create a dictionary using the list elements as keys, you ended up with a dictionary with a single element - because all of the list elements were identical.

要按照您的意图创建 10 个不同的占位符张量，您应该执行以下操作:

To create 10 different placeholder tensors, as you intended, you should instead do the following:

inputs = [tf.placeholder(tf.float32, shape=(batch_size, input_size))
          for _ in xrange(10)]

如果您打印此列表的元素，您会看到每个元素都是一个具有不同名称的张量.

If you print the elements of this list, you'll see that each element is a tensor with a different name.

^* 您现在可以将 元组 作为 feed_dict 的键传递，因为它们可能是用作字典键.

^* You can now pass tuples as the keys of a feed_dict, because these may be used as dictionary keys.

这篇关于将列表输入 TensorFlow 中的 feed_dict 时发出问题的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！