Yang An

Hi Dhruv, is there an update on the workaround? Thanks again for your time!

You can find a simple fix here: https://github.com/subhadarship/kmeans_pytorch/issues/16

You need to use the GitHub version, not the pip version. There is a flag called tqdm_flag. Set it to False: e.g. ` cluster_ids_x, cluster_centers = kmeans(X=x, num_clusters=num_clusters, distance='euclidean', device=torch.device('cuda:0'),...

For the reported results, we only used augment_data. We did not use augment_eth_ucy_social, it was an experimental version that didn't do what we expected.

The template reflects a probability distribution over positions, so everything should sum up to 1. As you mentioned, usually BCE is used for binary labels, but it can also be...