state_dict load error when evaluating

[BooPatienze]

I want to evaluate a SPECIFIED Image using your method, how should I do?

1. After downloading two file "swiftnet_pyr_cs.pt" and "swiftnet_ss_cs.pt" from the link https://drive.google.com/drive/folders/1DqX-N-nMtGG9QfMY_cKtULCKTfEuV4WT you give in README.md

2. I edit the file configs/rn18_single_scale.py to switch the evaluating to True and load the pre-trained model in this way:

67 if evaluating:
68    model.load_state_dict(torch.load('weights/swiftnet_ss_cs.pt'))
69     # model.load_state_dict(torch.load('weights/rn18_single_scale/model_best.pt'))
70 else:
71     model.criterion = SemsegCrossEntropy(num_classes=num_classes, ignore_id=ignore_id)
72     lr = 4e-4
73     lr_min = 1e-6
74     fine_tune_factor = 4
75     weight_decay = 1e-4
76     epochs = 250

3. then run this cmd for evaluation: python eval.py configs/rn18_single_scale.py but I got some state_dict errors:

RuntimeError: Error(s) in loading state_dict for SemsegModel:
        Missing key(s) in state_dict: "backbone.img_mean", "backbone.img_std", "logits.conv.bias".
        size mismatch for logits.conv.weight: copying a param with shape torch.Size([19, 128, 3, 3]) from checkpoint, the shape in current model is torch.Size([19, 128, 1, 1]).

4. So I change the pre-trained model I ever load in configs/rn18_single_scale.py from swiftnet_ss_cs.pt to swiftnet_pyr_cs.pt and then run last step: 3, I still got some state_dict error:

RuntimeError: Error(s) in loading state_dict for SemsegModel:
        Missing key(s) in state_dict:
        ……
        Unexpected key(s) in state_dict:
        ……
        size mismatch for logits.conv.weight: copying a param with shape torch.Size([19, 128, 3, 3]) from checkpoint, the shape in current model is torch.Size([19, 128, 1, 1]).

So, how should I do to evaluate a Specified Image?

[isunLt]

Same error