search

zhengye1995 / Zero-shot-Instance-Segmentation

code for CVPR paper Zero-shot Instance Segmentation
Apache License 2.0
93 stars 14 forks source link

RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [300, 49]] is at version 4; expected version 3 instead #30

Closed WindFall-wyx closed 2 years ago

WindFall-wyx commented 2 years ago

cls_score, bbox_pred, x_semantic, d_feature = self.bbox_head(bbox_feats, bg_vector=bg_vector[0].view([300,])) sum_cls_score, sum_bbox_pred,sum_x_semantic,sum_d_feature = self.bbox_head(sum_feats, bg_vector=bg_vector[0].view([300, ])) RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [300, 49]] is at version 4; expected version 3 instead. Hint: enable anomaly detection to find the operation that failed to compute its gradient, with torch.autograd.set_detect_anomaly(True). I used your bbox_head twice, it got this error. If with_syn_bg is closed, it functions well. How should i do to use it with syn_bg?

zhengye1995 commented 2 years ago

I think this should be a problem of pytorch itself, you can find the discussion in here.