debidatta

There was another issue where someone pointed out this is happening on Windows but not on Linux machines. Are you using Windows for this?

Yes a couple of changes are needed based on this [commit](https://github.com/google-research/google-research/commit/597136fbfd0f5fe7da2db811f662b3723dec1c81) Architecturally, pos_encoding1 and pos_encoding2 [here](https://github.com/materight/RepNet-pytorch/blob/main/repnet/model.py#L164) would now have 512 channels and fc_layer size for period prediction [here](https://github.com/materight/RepNet-pytorch/blob/main/repnet/model.py#L45) will have...