Incompatible input shape

[edmundngu]

May I know why i'm getting this error?
I believe that all the layers in my model are set correctly because I can train it without any problem.
However, the problem below occur when I test my model.
FYI, the problem arise when the model concat an upsampling layer with a conv layer.
Any solution?

Thanks to your help.

MXNetError: MXNetError: Error in operator RF_core_concat32: [05:38:05] ../src/operator/nn/concat.cc:67: Check failed: shape_assign(&(*in_shape)[i], dshape): Incompatible input shape: expected [1,-1,32,47], got [1,256,32,48]

[szha]

Looks like there's a mismatch in the size of the last dimension in the infer_shape of RF_core_concat32.

Incompatible input shape: expected [1,-1,32,47], got [1,256,32,48]

The operator doesn't seem to be from mxnet. Are you using custom op?

[edmundngu]

Thanks for your help @szha .
I found the solution for this problem.
I use the function below to crop the shape in order to match them up.

mx.symbol.Crop