Trying to concatenate two keras tensors but getting error

Question

I'm trying to concatenate a tensor with shape (None, 11, 1, 1) with another tensor with shape (None,1,1,1). Reading the keras.layers.Concatenate() docs I understood that the shape should be the same except for the concatenation axis which in this case would be axis 1. However when running the following code

import keras

mainInputShape = (11,1,1)
weightInputShape = (1,1,1)
mIn = keras.layers.Input(shape=mainInputShape, name='mainInput')
wIn = keras.layers.Input(shape=weightInputShape, name='weightInput') 
x = keras.layers.Concatenate(axis=0)([mIn, wIn])

I get the following error

  Cell In[10], line 7
    x = keras.layers.Concatenate(axis=1)([mIn, wIn])

  File ~\anaconda3\Lib\site-packages\keras\src\utils\traceback_utils.py:122 in error_handler
    raise e.with_traceback(filtered_tb) from None

  File ~\anaconda3\Lib\site-packages\keras\src\layers\merging\concatenate.py:70 in build
    del reduced_inputs_shapes[i][axis]

IndexError: list assignment index out of range

Could someone please explain what's happening and how to fix this? I need an output tensor of shape (None, 12,1,1)?

Hello! I can't reproduce your error. Could you specify your keras version and your tensorflow version? Thanks — Lescurel
– Lescurel, Commented Jun 20, 2024 at 12:21

Balaji Ambresh · Accepted Answer · 2024-06-21 15:05:41Z

0

Tensorflow is buggy when handling this case (report here).

>>> import numpy as np
>>> x = np.random.random((32, 11, 1, 1))
>>> y = np.random.random((32, 1, 1, 1))
>>> np.concatenate([x, y], axis=1).shape
(32, 12, 1, 1)

answered Jun 21, 2024 at 15:05

Balaji Ambresh

5,0022 gold badges7 silver badges17 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

Trying to concatenate two keras tensors but getting error

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related