A checker is needed for inputs of Conv layers.

[Shuo-Sun20]

Issue type

Bug

Have you reproduced the bug with TensorFlow Nightly?

Yes

Source

source

TensorFlow version

tf 2.16

Custom code

Yes

OS platform and distribution

WIndows10 & Linux

Mobile device

No response

Python version

No response

Bazel version

No response

GCC/compiler version

No response

CUDA/cuDNN version

No response

GPU model and memory

No response

Current behavior?

As disscussed in #64396, it should be considered as an abnormal input when kernel_size > input_size, while currently the Conv layers will silently generate an unexpected output, therefore maybe a checker is needed here.

The code to reproduce this issue:

from keras.layers import Conv2D
import numpy as np

x=np.random.rand(1,2,2,1)
l=Conv2D(1,3,(1,1),'valid','channels_last', [1,1],1, 'linear', True)
print(l(x).shape)
print(l.compute_output_shape(x.shape))

In my opinion, this code should not generate an output, a warning or error is expected since the kernel(3x3) is larger than input (2x2).

Standalone code to reproduce the issue

from keras.layers import Conv2D
import numpy as np

x=np.random.rand(1,2,2,1)
l=Conv2D(1,3,(1,1),'valid','channels_last', [1,1],1, 'linear', True)
print(l(x).shape)
print(l.compute_output_shape(x.shape))

Relevant log output

Currently in windows, the output is
(1, 2, 2, 1)
(1, 0, 0, 1)
And in Linux the output is
(1, 0, 0, 1)
(1, 0, 0, 1)

The inconsistant is discussed in tensorflow/tensorflow#64396. This issue is to report the need of checker for abnormal inputs of Conv layers.

[Venkat6871]

Hi @Shuo-Sun20 ,

I tried to run your code on colab using TF-nightly and faced the same issue. Please find the gist here for reference.

Thank you!