search

zhengye1995 / Zero-shot-Instance-Segmentation

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

What happend when I ran ./tools/dist_test.sh #20

Closed algoteam5 closed 3 years ago

algoteam5 commented 3 years ago

What does the following output indicate when I ran this command ./tools/dist_test.sh configs/zsi/65_15/test/gzsi/zero-shot-mask-rcnn-BARPN-bbox_mask_sync_bg_65_15_decoder_notanh_gzsi.py checkpoints/ZSI_65_15.pth 4 --show?


gx = torch.addcmul(px, 1, pw, dx)  # gx = px + pw * dx
[>>>>>>>>>>>>>>>>>>>>>>>>>>>>>          ] 8112/10098, 12.9 task/s, elapsed: 654s, ETA
zhengye1995 commented 3 years ago

Nothing happend except the inference process because --show only works in single gpu model while you use 4 gpu to inference. See code here.

algoteam5 commented 3 years ago

@zhengye1995 Why the inference took so long? normally it requires a few milliseconds to get inference from a trained model.

zhengye1995 commented 3 years ago

@algoteam5 Because this process inferences all 10098 images in test label file, which is determined by this code in config file.