occu_mask
lt573152873 opened this issue · comments
lt573152873 commented
during training, Occu_mask whether there is gradient return? when occu_mask is zeros, the loss is small;
if need occu_mask -> occu_mask.detach() ?
The official PyTorch implementation of the paper "Learning by Analogy: Reliable Supervision from Transformations for Unsupervised Optical Flow Estimation".
lt573152873 opened this issue · comments
during training, Occu_mask whether there is gradient return? when occu_mask is zeros, the loss is small;
if need occu_mask -> occu_mask.detach() ?