mxnet>1.4.1 returns error for the deconvolution function

[lbcg]

Hello @chinakook , thanks for the good work here.
I have encountered the following error working with mxnet==1.5.0:

mxnet.base.MXNetError: Error in operator dilatedunet0_conv6_deconvolution0: [11:32:34] C:\Jenkins\workspace\mxnet-tag\mxnet\src\operator\nn\deconvolution.cc:131: Check failed: dshape_nchw[1] % param_.num_group == 0U (47 vs. 0) : input num_filter must divide group size

environment

anaconda with python 3.6.4
pip install mxnet>=1.5.0
pycharm 2019.1.1
windows

Steps to reproduce error

python model_dilated_unet.py

For error message, see traceback
traceback unet.txt

resolution

pip uninstall mxnet
pip install mxnet==1.4.1

Since this mxnet version is now outdated, you might want to find a way to make your code compatible with newer mxnet versions.

[chinakook]

Deconvolution is not dropped, I'll upload a new model later.

[jens-maus]

I am facing the same problem here since having updated to mxnet >= 1.5.0. So can you please update your model accordingly so that the error message for the deconvolution disappears. Thanks.

[thimabru1010]

Still having the same error. When you @chinakook will fix?

[handong1587]

Does anyone solve this problem?