Appllying strict=False for using a modified pretrained model

1

I want to remove and modify the code of a preteined model and for loading this pretrained model I want to use strict=False in the state_dict. Based on the code I’m working with and also based on the pretrained model that I need to use ([TimeSformer](https://github.com/facebookresearch/TimeSformer)), the TimeSformer model load its weights from inside itself.but in my previous experiments (using the current codes) I loaded the pretrained weights in another section of the code. In my current code, I cannot find any place to apply strict=False . How can I apply strict=False to avoid any mismatch because of differences between modified code of the pretrained model and its weights? Any help would be appreciated.

This is my code that I used for my previous experiments:

class my_model(nn.Module):
    def __init__(self, pretrained=False):
        super(my_model, self).__init__()
        
        
        self.featureExtractor =feature_extractor()
        
        
        if pretrained:
            print('Loading weights...')
            weight_dict=torch.load(os.path.join('models','vid_class.pt'))

            model_dict=self.featureExtractor.state_dict()
            
            list_weight_dict=list(weight_dict.items())
            list_model_dict=list(model_dict.items())
            
            for i in range(len(list_model_dict)):
                assert list_model_dict[i][1].shape==list_weight_dict[i][1].shape
                model_dict[list_model_dict[i][0]].copy_(weight_dict[list_weight_dict[i][0]])
            
            for i in range(len(list_model_dict)):
                assert torch.all(torch.eq(model_dict[list_model_dict[i][0]],weight_dict[list_weight_dict[i][0]].to('cpu')))
            print('Loading done!')

But when using TimeSformer, it loads the weights by itself and cannot find the place for applying strict=False in the state_dict:

class my_model(nn.Module):
    def __init__(self, pretrained=False):
        super(my_model, self).__init__()
        self.featureExtractor =TimeSformer(img_size=224, num_classes=400, num_frames=8, attention_type='divided_space_time',  
                                           pretrained_model='/models/TimeSformer.pyth')
..
..
..