KeyError: 'base.conv1.weight'

[x1ng4me]

Trying to run the code. Got the error

KeyError: 'base.conv1.weight' in backbone file.

Have any idea to solve this?

Thank you.

[HWNHJJ]

多GPU训练保存的参数单GPU的环境无法直接用

[mojiayuan]

多GPU训练保存的参数单GPU的环境无法直接用
您好，这个解决不了吗

[DoranLyong]

As I can't use Baidu account, I downloaded the pretrained model from here.
Then, I got the same problem.

I found out that the pretrained model file has different layer names with this repository.

So, I fixed the code block a bit and it worked!

Go to .../model/backbones/resnet.py
and I changed def load_param(self, model_path) block like:

    def load_param(self, model_path):
        param_dict = torch.load(model_path)
        for i in param_dict:
            j = i.replace("base.","")
            if 'fc' in i:
                continue
            if j in self.state_dict().keys():
                self.state_dict()[j].copy_(param_dict[i])