ValueError: Operands could not be broadcast together with shapes , when converting with change_ordering

[shairoz-deci]

Describe the bug
When loading a model and trying to convert it from pytorch to keras with change_ordering=True, I get an error.
is there a way to resolve or figure out which layers require tuning?
Running with change_ordering=False works properly

  File "/home/shai/.conda/envs/shai-development/lib/python3.7/site-packages/tensorflow/python/keras/layers/merge.py", line 85, in _compute_elemwise_op_output_shape
    'together with shapes ' + str(shape1) + ' ' + str(shape2))
ValueError: Operands could not be broadcast together with shapes (56, 56, 64) (64, 1, 64)

To Reproduce

from timm.models.regnet import regnety_008
import pytorch2keras
dims = (1, 3, 224, 224)
keras_model = pytorch2keras.pytorch_to_keras(regnety_008(), torch.randn(dims, requires_grad=False), [dims[1:]], verbose=True, change_ordering=True)

Environment (please complete the following information):

• OS: [e.g. iOS]
• Python 3.7
• TF 2.4.0
• torch 1.7.1
• pytorch2keras 0.2.4

[robertarnesson]

Any luck? Having the same problem