lm_head.weight in convert_opt_checkpoint.py

[quaternior]

Hello authors, I would like to express my gratitude for thecontributions you've made.
I read convert_opt_checkpoint.py, and I found the code as below.

    item['lm_head.weight'] = model.state_dict()['model.decoder.embed_tokens.weight']
    item['final_layer_norm.weight'] = model.state_dict()['model.decoder.final_layer_norm.weight']
    item['final_layer_norm.bias'] = model.state_dict()['model.decoder.final_layer_norm.bias']

Since the key is 'lm_head.weight', shouldn't it be read as the key for 'lm_head.weight' in the model as well?
Thanks!