Reproducing results on AMASS

[meaten]

Hi,

Thanks for your great work!

I'm trying to reproduce the AMASS results. I successfully reproduced the results using BeLFusion with the pre-trained autoencoder and BLS.
However, I encountered issues when training the autoencoder and BLS from scratch.
The ADE and FDE are worse by around 0.5.

It would be incredibly helpful if you could provide the info.log file for the autoencoder and BLS on the AMASS dataset.