Question

我正在尝试在Tensorflow 2.0RC中实现自定义Keras Layer，并且需要将[None, Q]形张量连接到[None, H, W, D]形张量上以产生[None, H, W, D + Q]形张量。假定两个输入张量具有相同的批量大小，即使事先未知。同样，H，W，D和Q在写入时都不是已知的，但是在首次调用该层时，会在该层的build方法中对其进行评估。我遇到的问题是将[None, Q]形张量传播到[None, H, W, Q]形张量以进行串联时。

以下是尝试使用功能性API创建Keras Model的示例，该API执行从形状[None, 3]到形状[None, 5, 5, 3]的可变批量广播：

import tensorflow as tf
import tensorflow.keras.layers as kl
import numpy as np

x = tf.keras.Input([3])  # Shape [None, 3]
y = kl.Reshape([1, 1, 3])(x)  # Need to add empty dims before broadcasting
y = tf.broadcast_to(y, [-1, 5, 5, 3])  # Broadcast to shape [None, 5, 5, 3]

model = tf.keras.Model(inputs=x, outputs=y)

print(model(np.random.random(size=(8, 3))).shape)

Tensorflow产生错误：

InvalidArgumentError:  Dimension -1 must be >= 0

然后，当我将-1更改为None时，它会给我：

TypeError: Failed to convert object of type <class 'list'> to Tensor. Contents: [None, 5, 5, 3]. Consider casting elements to a supported type.

如何执行指定的广播？

Answer 1

您需要使用y的动态形状来确定批量大小。张量y的动态形状由tf.shape(y)给出，它是一个张量op，表示在运行时评估的y的形状。修改后的示例通过使用tf.where在旧形状[None, 1, 1, 3]和新形状之间进行选择来说明这一点。

import tensorflow as tf
import tensorflow.keras.layers as kl
import numpy as np

x = tf.keras.Input([3])  # Shape [None, 3]
y = kl.Reshape([1, 1, 3])(x)  # Need to add empty dims before broadcasting
# Retain the batch and depth dimensions, but broadcast along H and W
broadcast_shape = tf.where([True, False, False, True],
                           tf.shape(y), [0, 5, 5, 0])
y = tf.broadcast_to(y, broadcast_shape)  # Broadcast to shape [None, 5, 5, 3]

model = tf.keras.Model(inputs=x, outputs=y)

print(model(np.random.random(size=(8, 3))).shape)
# prints: "(8, 5, 5, 3)"

参考文献：

"TensorFlow: Shapes and dynamic dimensions"

显式广播可变批量大小张量

1 个答案: