TypeError: Layer reshape_1 does not support masking, but was passed an input_mask: Tensor("concatena

原创于 2019-08-21 20:33:52 发布 · 3.5k 阅读

0 ·

CC 4.0 BY-SA版权

人工智能专栏收录该内容

5 篇文章

订阅专栏

本文详细探讨了在使用Keras框架时遇到的Reshape层错误，并提供了一个有效的解决方案。通过插入Lambda层，可以避免Reshape层不支持masking的问题。此外，文章邀请读者分享更多关于此问题的操作技巧。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

Keras Reshape层报错

Traceback (most recent call last):
    pcsel = Reshape((-1, 20, 30))(pcsel)
  File "/home/xxx/anaconda3/envs/python36/lib/python3.6/site-packages/keras/engine/base_layer.py", line 458, in __call__
    output_mask = self.compute_mask(inputs, previous_mask)
  File "/home/xxx/anaconda3/envs/python36/lib/python3.6/site-packages/keras/engine/base_layer.py", line 616, in compute_mask
    str(mask))
TypeError: Layer reshape_1 does not support masking, but was passed an input_mask: Tensor("concatenate_1/All:0", shape=(?, ?), dtype=bool)

解决方案:

I found it also works if you insert a Lambda layer before Reshape like this:
x = Lambda(lambda x: x, output_shape=lambda s:s)(x).

But the best thing I guess is to make some operations at this layer. Anyone has a good idea please shows here.

参考链接 https://github.com/keras-team/keras/issues/4978