The model and loaded state dict do not match exactly

[Zhcyoung]

I pretrained a self-supervised model (BEiT) using my own data, but encountered an error after using ./tools/model_converters/extract_backbone_weights.py extract the backbone. This error still exists when I use the given imageNet weights. Is this a normal phenomenon？

The model and loaded state dict do not match exactly

unexpected key in source state_dict: backbone.mask_token, 
backbone.rel_pos_bias.relative_position_bias_table, backbone.rel_pos_bias.relative_position_index

missing keys in source state_dict: backbone.layers.0.attn.relative_position_bias_table, backbone.layers.0.attn.relative_position_index, 
backbone.layers.1.attn.relative_position_bias_table, backbone.layers.1.attn.relative_position_index, 
backbone.layers.2.attn.relative_position_bias_table, backbone.layers.2.attn.relative_position_index, 
backbone.layers.3.attn.relative_position_bias_table, backbone.layers.3.attn.relative_position_index, 
backbone.layers.4.attn.relative_position_bias_table, backbone.layers.4.attn.relative_position_index, 
backbone.layers.5.attn.relative_position_bias_table, backbone.layers.5.attn.relative_position_index, 
backbone.layers.6.attn.relative_position_bias_table, backbone.layers.6.attn.relative_position_index, 
backbone.layers.7.attn.relative_position_bias_table, backbone.layers.7.attn.relative_position_index, 
backbone.layers.8.attn.relative_position_bias_table, backbone.layers.8.attn.relative_position_index, 
backbone.layers.9.attn.relative_position_bias_table, backbone.layers.9.attn.relative_position_index, 
backbone.layers.10.attn.relative_position_bias_table, backbone.layers.10.attn.relative_position_index, 
backbone.layers.11.attn.relative_position_bias_table, backbone.layers.11.attn.relative_position_index, backbone.ln2.weight, 
backbone.ln2.bias, head.fc.weight, head.fc.bias