Open dorsadadjoo opened 3 years ago
To solve the previous issue, I added the strict=False attribute to load_state_dict() (File "evaluate_cub.py", line 135) :
model.load_state_dict(torch.load(args.restore_from), strict=False)
I also changed my torch version to 1.4.0.
Then, I got a file extension error while running sh eval_cub.sh and the reason was that the saving path in Image.fromarray was hardcoded (File "evaluate_cub.py", line 244) filename path was local and hardcoded. To solve that, the img_folder (line 190) and img_nm (line 192) should change depending on your local path.
Thanks this worked for me
I ran eval_cub.sh with model=DeepLab50_2branch
and got the following error:
Also, I tried other models. Seems like there is a mismatch between the model and the provided snapshot.