About loading pre-training errors

[linyi785]

First of all, thank you for your code, which has benefited me a lot, but the same error always occurs when loading the pre-training model. The error code is as follows
Error(s) in loading state_dict for OctResNet:Missing key(s) in state_dict:
Can you help me with it, or can you provide me with a full version of the code? I am grateful! My email is bruceleely@outlook.com

[d-li14]

@linyi785 Could you describe the detailed errors?
Everything went well on my side following these simple lines, without mismatched keys.

import torch
from oct_resnet import oct_resnet50
net = oct_resnet50()
net.load_state_dict(torch.load('/path/to/oct_resnet50_cosine.pth'))

[linyi785]

RuntimeError: Error(s) in loading state_dict for OctResNet:
Missing key(s) in state_dict: "layer1.0.conv1.bn_h.weight", "layer1.0.conv1.bn_h.bias", "layer1.0.conv1.bn_h.running_mean", "layer1.0.conv1.bn_h.running_var", "layer1.0.conv1.bn_l.weight", "layer1.0.conv1.bn_l.bias", "layer1.0.conv1.bn_l.running_mean", "layer1.0.conv1.bn_l.running_var", "layer1.0.conv2.bn_h.weight", "layer1.0.conv2.bn_h.bias", "layer1.0.conv2.bn_h.running_mean", "layer1.0.conv2.bn_h.running_var", "layer1.0.conv2.bn_l.weight", "layer1.0.conv2.bn_l.bias", "layer1.0.conv2.bn_l.running_mean", "layer1.0.conv2.bn_l.running_var", "layer1.0.conv3.conv.conv_l2l.weight", "layer1.0.conv3.conv.conv_l2h.weight", "layer1.0.conv3.conv.conv_h2l.weight", "layer1.0.conv3.conv.conv_h2h.weight", "layer1.0.conv3.bn_h.weight", "layer1.0.conv3.bn_h.bias", "layer1.0.conv3.bn_h.running_mean", "layer1.0.conv3.bn_h.running_var", "layer1.0.conv3.bn_l.weight", "layer1.0.conv3.bn_l.bias", "layer1.0.conv3.bn_l.running_mean", "layer1.0.conv3.bn_l.running_var", "layer1.0.downsample.0.conv.conv_h2l.weight", "layer1.0.downsample.0.conv.conv_h2h.weight", "layer1.0.downsample.0.bn_h.weight", "layer1.0.downsample.0.bn_h.bias", "layer1.0.downsample.0.bn_h.running_mean", "layer1.0.downsample.0.bn_h.running_var", "layer1.0.downsample.0.bn_l.weight", "layer1.0.downsample.0.bn_l.bias", "layer1.0.downsample.0.bn_l.running_mean", "layer1.0.downsample.0.bn_l.running_var", "layer1.1.conv1.bn_h.weight", "layer1.1.conv1.bn_h.bias", "layer1.1.conv1.bn_h.running_mean", "layer1.1.conv1.bn_h.running_var", "...
Unexpected key(s) in state_dict: "layer1.0.bn1.running_mean", "layer1.0.bn1.running_var", "layer1.0.bn1.weight", "layer1.0.bn1.bias", "layer1.0.bn2.running_mean", "layer1.0.bn2.running_var", "layer1.0.bn2.weight", "layer1.0.bn2.bias", "layer1.0.bn3.running_mean", "layer1.0.bn3.running_var", "layer1.0.bn3.weight", "layer1.0.bn3.bias", "layer1.0.conv1.weight", "layer1.0.conv2.weight", "layer1.0.conv3.weight", "layer1.0.downsample.1.running_mean", "layer1.0.downsample.1.running_var", "layer1.0.downsample.1.weight", "layer1.0.downsample.1.bias", "layer1.0.downsample.0.weight", "layer1.1.bn1.running_mean", "layer1.1.bn1.running_var", "layer1.1.bn1.weight", "layer1.1.bn1.bias", "layer1.1.bn2.running_mean", "layer1.1.bn2.running_var", "layer1.1.bn2.weight", "layer1.1.bn2.bias", "layer1.1.bn3.running_mean", "layer1.1.bn3.running_var", "layer1.1.bn3.weight", "layer1.1.bn3.bias", "layer1.1.conv1.weight", "layer1.1.conv2.weight", "layer1.1.conv3.weight", "layer1.2.bn1.running_mean", "layer1.2.bn1.running_var", "layer1.2.bn1.weight", "layer1.2.bn1.bias", "layer1.2.bn2.running_mean", "layer1.2.bn2.running_var", "layer1.2.bn2.weight", "layer1.2.bn2.bias", "layer1.2.bn3.running_mean", "layer1.2.bn3.running_var", "layer1.2.bn3.weight", "layer1.2.bn3.bias", "layer1.2.conv1.weight", "layer1.2.conv2.weight", "layer1.2.conv3.weight", "layer2.0.bn1.running_mean", "layer2.0.bn1.running_var", "layer2.0.bn1.weight", "layer2.0.bn1.bias", "layer2.0.bn2.running_mean", "layer2.0.bn2.running_var", "l...

[linyi785]

The pre-training weight I used is that the standard resnet50 is not oct_resnet50. Is it wrong here？

[d-li14]

The architecture of vanilla ResNet and OctResNet is different in almost all layers, so directly loading the weight of ResNet to OctResNet causes the error you met.

[linyi785]

thanks！By the way，Is your pre-training weight trained on the ImageNet dataset?

[d-li14]

yes

[linyi785]

Thank you very much for solving all my doubts.