2020-11-20

RuntimeError: Error(s) in loading state_dict for DABNet:
    Missing key(s) in state_dict: "init_conv.0.conv.weight", "init_conv.0.bn_prelu.bn.weight", "init_conv.0.bn_prelu.bn.bias","classifier.0.conv.weight".
    Unexpected key(s) in state_dict: "epoch", "model".等一大堆的错误

源码：

import torch
from model.DABNet import DABNet

model_dir = "./all_model_dir/dataset/mic/DABNetbs20gpu1_trainval/model_100.pth"
example = torch.rand(1,3,400,400)

net = DABNet(classes=2)
state_dict = torch.load(model_dir, map_location='cpu')
net.load_state_dict(state_dict)
net.eval()
traced_script_module = torch.jit.trace(net, example)
traced_script_module.save("traced_DABnet_model.pt")

报错原因：

是因为训练环境和到处环境不一致

更改　

net.load_state_dict(state_dict)　

为　

net.load_state_dict(state_dict，FALSE)

即可解决