Errors in loading state_dict (size mismatch)

[dubdu]

Hey, can you please help me?

I was trying to train the model on custom dataset following instructions from this issue (deleted LPRDataLoader.getitem and modified CHARS, though did not change T_length).

I am running the code from Google Colaboratory and here is the issue:

Successful to build network!

RuntimeError Traceback (most recent call last)
in ()
154
155 if name == "main":
--> 156 test()

1 frames
/usr/local/lib/python3.7/dist-packages/torch/nn/modules/module.py in load_state_dict(self, state_dict, strict)
1496 if len(error_msgs) > 0:
1497 raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
-> 1498 self.class.name, "\n\t".join(error_msgs)))
1499 return _IncompatibleKeys(missing_keys, unexpected_keys)
1500

RuntimeError: Error(s) in loading state_dict for LPRNet:
size mismatch for backbone.20.weight: copying a param with shape torch.Size([68, 256, 13, 1]) from checkpoint, the shape in current model is torch.Size([37, 256, 13, 1]).
size mismatch for backbone.20.bias: copying a param with shape torch.Size([68]) from checkpoint, the shape in current model is torch.Size([37]).
size mismatch for backbone.21.weight: copying a param with shape torch.Size([68]) from checkpoint, the shape in current model is torch.Size([37]).
size mismatch for backbone.21.bias: copying a param with shape torch.Size([68]) from checkpoint, the shape in current model is torch.Size([37]).
size mismatch for backbone.21.running_mean: copying a param with shape torch.Size([68]) from checkpoint, the shape in current model is torch.Size([37]).
size mismatch for backbone.21.running_var: copying a param with shape torch.Size([68]) from checkpoint, the shape in current model is torch.Size([37]).
size mismatch for container.0.weight: copying a param with shape torch.Size([68, 516, 1, 1]) from checkpoint, the shape in current model is torch.Size([37, 485, 1, 1]).
size mismatch for container.0.bias: copying a param with shape torch.Size([68]) from checkpoint, the shape in current model is torch.Size([37]).

Code itself:

def test():
args = get_parser()
lprnet = build_lprnet(lpr_max_len=args.lpr_max_len, phase=args.phase_train, class_num=len(CHARS), dropout_rate=args.dropout_rate)
device = torch.device("cuda:0" if args.cuda else "cpu")
lprnet.to(device)
print("Successful to build network!")

# load pretrained model
if args.pretrained_model:
    lprnet.load_state_dict(torch.load(args.pretrained_model)) _##underlined_
    print("load pretrained model successful!")
else:
    print("[Error] Can't found pretrained mode, please check!")
    return False
test_img_dirs = os.path.expanduser(args.test_img_dirs)
test_dataset = LPRDataLoader(test_img_dirs.split(','), args.img_size, args.lpr_max_len)
try:
    Greedy_Decode_Eval(lprnet, test_dataset, args)
finally:
    cv2.destroyAllWindows()

What should I do?..

[Vergissmeinnic]

I'm the same with you.

[MyraBaba]

same error:

Successful to build network! Traceback (most recent call last): File "/Applications/PyCharm CE.app/Contents/plugins/python-ce/helpers/pydev/pydevd.py", line 1483, in _exec pydev_imports.execfile(file, globals, locals) # execute the script File "/Applications/PyCharm CE.app/Contents/plugins/python-ce/helpers/pydev/_pydev_imps/_pydev_execfile.py", line 18, in execfile exec(compile(contents+"\n", file, 'exec'), glob, loc) File "/Users/tulpar/Project/LPRNet_Pytorch/test_LPRNet.py", line 205, in <module> test() File "/Users/tulpar/Project/LPRNet_Pytorch/test_LPRNet.py", line 73, in test lprnet.load_state_dict(torch.load(args.pretrained_model, map_location=torch.device(device))) File "/Users/tulpar/Projects/Deepfake-Detection2/venv/InsightFace-REST-Nov2021/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1497, in load_state_dict raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format( RuntimeError: Error(s) in loading state_dict for LPRNet: Missing key(s) in state_dict: "stage1.0.weight", "stage1.0.bias", "stage1.1.weight", "stage1.1.bias", "stage1.1.running_mean", "stage1.1.running_var", "stage1.3.block.0.weight", "stage1.3.block.0.bias", "stage1.3.block.1.weight", "stage1.3.block.1.bias", "stage1.3.block.1.running_mean", "stage1.3.block.1.running_var", "stage1.3.block.3.weight", "stage1.3.block.3.bias", "stage1.3.block.4.weight", "stage1.3.block.4.bias", "stage1.3.block.4.running_mean", "stage1.3.block.4.running_var", "stage1.4.downsample.block.0.weight", "stage1.4.downsample.block.0.bias", "stage1.4.block.0.weight", "stage1.4.block.0.bias", "stage1.4.block.1.weight", "stage1.4.block.1.bias", "stage1.4.block.1.running_mean", "stage1.4.block.1.running_var", "stage1.4.block.3.weight", "stage1.4.block.3.bias", "stage1.4.block.4.weight", "stage1.4.block.4.bias", "stage1.4.block.4.running_mean", "stage1.4.block.4.running_var", "stage1.5.downsample.block.0.weight", "stage1.5.downsample.block.0.bias", "stage1.5.block.0.weight", "stage1.5.block.0.bias", "stage1.5.block.1.weight", "stage1.5.block.1.bias", "stage1.5.block.1.running_mean", "stage1.5.block.1.running_var", "stage1.5.block.3.weight", "stage1.5.block.3.bias", "stage1.5.block.4.weight", "stage1.5.block.4.bias", "stage1.5.block.4.running_mean", "stage1.5.block.4.running_var", "stage1.6.downsample.block.0.weight", "stage1.6.downsample.block.0.bias", "stage1.6.block.0.weight", "stage1.6.block.0.bias", "stage1.6.block.1.weight", "stage1.6.block.1.bias", "stage1.6.block.1.running_mean", "stage1.6.block.1.running_var", "stage1.6.block.3.weight", "stage1.6.block.3.bias", "stage1.6.block.4.weight", "stage1.6.block.4.bias", "stage1.6.block.4.running_mean", "stage1.6.block.4.running_var", "downsample1.0.weight", "downsample1.0.bias", "downsample1.1.weight", "downsample1.1.bias", "downsample1.1.running_mean", "downsample1.1.running_var", "downsample1.3.weight", "downsample1.3.bias", "downsample1.4.weight", "downsample1.4.bias", "downsample1.4.running_mean", "downsample1.4.running_var", "downsample1.6.weight", "downsample1.6.bias", "downsample1.7.weight", "downsample1.7.bias", "downsample1.7.running_mean", "downsample1.7.running_var", "stage2.0.downsample.block.0.weight", "stage2.0.downsample.block.0.bias", "stage2.0.block.0.weight", "stage2.0.block.0.bias", "stage2.0.block.1.weight", "stage2.0.block.1.bias", "stage2.0.block.1.running_mean", "stage2.0.block.1.running_var", "stage2.0.block.3.weight", "stage2.0.block.3.bias", "stage2.0.block.4.weight", "stage2.0.block.4.bias", "stage2.0.block.4.running_mean", "stage2.0.block.4.running_var", "stage2.1.block.0.weight", "stage2.1.block.0.bias", "stage2.1.block.1.weight", "stage2.1.block.1.bias", "stage2.1.block.1.running_mean", "stage2.1.block.1.running_var", "stage2.1.block.3.weight", "stage2.1.block.3.bias", "stage2.1.block.4.weight", "stage2.1.block.4.bias", "stage2.1.block.4.running_mean", "stage2.1.block.4.running_var", "downsample2.0.weight", "downsample2.0.bias", "downsample2.1.weight", "downsample2.1.bias", "downsample2.1.running_mean", "downsample2.1.running_var", "downsample2.3.weight", "downsample2.3.bias", "downsample2.4.weight", "downsample2.4.bias", "downsample2.4.running_mean", "downsample2.4.running_var", "stage3.0.downsample.block.0.weight", "stage3.0.downsample.block.0.bias", "stage3.0.block.0.weight", "stage3.0.block.0.bias", "stage3.0.block.1.weight", "stage3.0.block.1.bias", "stage3.0.block.1.running_mean", "stage3.0.block.1.running_var", "stage3.0.block.3.weight", "stage3.0.block.3.bias", "stage3.0.block.4.weight", "stage3.0.block.4.bias", "stage3.0.block.4.running_mean", "stage3.0.block.4.running_var", "stage3.1.downsample.block.0.weight", "stage3.1.downsample.block.0.bias", "stage3.1.block.0.weight", "stage3.1.block.0.bias", "stage3.1.block.1.weight", "stage3.1.block.1.bias", "stage3.1.block.1.running_mean", "stage3.1.block.1.running_var", "stage3.1.block.3.weight", "stage3.1.block.3.bias", "stage3.1.block.4.weight", "stage3.1.block.4.bias", "stage3.1.block.4.running_mean", "stage3.1.block.4.running_var", "stage4.0.weight", "stage4.0.bias", "stage4.1.weight", "stage4.1.bias", "stage4.1.running_mean", "stage4.1.running_var", "stage4.3.weight", "stage4.3.bias", "stage4.4.weight", "stage4.4.bias", "stage4.4.running_mean", "stage4.4.running_var", "bn.0.weight", "bn.0.bias", "bn.0.running_mean", "bn.0.running_var", "bn4.0.weight", "bn4.0.bias", "bn4.0.running_mean", "bn4.0.running_var". Unexpected key(s) in state_dict: "backbone.0.weight", "backbone.0.bias", "backbone.1.weight", "backbone.1.bias", "backbone.1.running_mean", "backbone.1.running_var", "backbone.1.num_batches_tracked", "backbone.4.block.0.weight", "backbone.4.block.0.bias", "backbone.4.block.2.weight", "backbone.4.block.2.bias", "backbone.4.block.4.weight", "backbone.4.block.4.bias", "backbone.4.block.6.weight", "backbone.4.block.6.bias", "backbone.5.weight", "backbone.5.bias", "backbone.5.running_mean", "backbone.5.running_var", "backbone.5.num_batches_tracked", "backbone.8.block.0.weight", "backbone.8.block.0.bias", "backbone.8.block.2.weight", "backbone.8.block.2.bias", "backbone.8.block.4.weight", "backbone.8.block.4.bias", "backbone.8.block.6.weight", "backbone.8.block.6.bias", "backbone.9.weight", "backbone.9.bias", "backbone.9.running_mean", "backbone.9.running_var", "backbone.9.num_batches_tracked", "backbone.11.block.0.weight", "backbone.11.block.0.bias", "backbone.11.block.2.weight", "backbone.11.block.2.bias", "backbone.11.block.4.weight", "backbone.11.block.4.bias", "backbone.11.block.6.weight", "backbone.11.block.6.bias", "backbone.12.weight", "backbone.12.bias", "backbone.12.running_mean", "backbone.12.running_var", "backbone.12.num_batches_tracked", "backbone.16.weight", "backbone.16.bias", "backbone.17.weight", "backbone.17.bias", "backbone.17.running_mean", "backbone.17.running_var", "backbone.17.num_batches_tracked", "backbone.20.weight", "backbone.20.bias", "backbone.21.weight", "backbone.21.bias", "backbone.21.running_mean", "backbone.21.running_var", "backbone.21.num_batches_tracked". size mismatch for container.0.weight: copying a param with shape torch.Size([68, 516, 1, 1]) from checkpoint, the shape in current model is torch.Size([11, 779, 1, 1]). size mismatch for container.0.bias: copying a param with shape torch.Size([68]) from checkpoint, the shape in current model is torch.Size([11]). python-BaseException

[mesakarghm]

This is probably because of the change in class_num parameter. If you changed the characters the model is going to train on, you cannot use the pretrained model. Better go with xavier initialization