Poor result with pre-trained model LIP_best.pth

[maguro27]

I also encountered the issue as well as liutinglt/CE2P#34.
However, this issue does not provide a solution.
I do not change the part of weight loading and it does not output any error messages.
How do I fix it?
Note that I referred to the CE2P repository (https://github.com/liutinglt/CE2P) so that exchange ABN for PyTorch ver 0.x to ABN for PyTorch ver 1.x.

The output of the network goes from about -0.02 to 0.03. Is it collect?
The network generates a poor result though I exchanged ABN to BN.
Thanks.

--My environment--
GPU: Tesla V100
PyTorch: 1.6
CUDA: 10.2.89
cuDNN: 8.0.2
ABN: https://github.com/mapillary/inplace_abn
Pre-trained model: LIP_best.pth

[ziwei-zh]

Did you test on LIP or on your own dataset?