Xuanlong Yu

icon indicating a website link xuanlong-yu.github.io

icon company Paris-Saclay University | ENSTA Paris icon location China icon location Ph.D. student at Paris-Saclay University and ENSTA Paris.

Results 3 comments of Xuanlong Yu

Nan loss during training

Hi, thank you for leaving the comments. Actually, it appears in the original bts training as well, you can check [here](https://github.com/cleinc/bts/issues/36) to find some suggestions given by the authors of...

Nan loss during training

Several common loss functions are indeed non-negative such as MSELoss or L1Loss. But it is fine to have a negative loss value during training, not only in LDU case. You...

Question about dissimilar_loss

Thank you for your comment. torch.cdist(a,a) calculates distance between _each_ pair of the two collections of row vectors. So it will output 0 on the diagonal of the output matrix,...