This is my code:

class EntityRetrieval(object):

    def __init__(self,
                 entity_kb_path: str,
                 save_after_init=True):

        """
        Args
            entity_kb_path: path of entity knowledge base (string) which is json format
        """
        
        try:
            entity_kb = json.load(open(entity_kb_path, 'r'))
        except Exception as e:
            logging.error(e)
            raise e 
            
        # Don't use entity type 
        entity_dict = {}
        for entity_type, entities in entity_kb.items():
            for entity_id, list_entity_strings in entities.items():
                entity_dict[entity_id] = list_entity_strings

        all_entity_strings = []
        all_entity_ids = []
        for entity_id, list_entity_strings in entity_dict.items():
            all_entity_strings.extend(list_entity_strings)
            all_entity_ids.extend([entity_id] * len(list_entity_strings))

        self.char_tfidf = TfidfVectorizer(analyzer='char',
                                          lowercase=True,
                                          ngram_range=(2, 5))

        self.char_tfidf.fit(all_entity_strings)
        self.entity_kb_features = self.char_tfidf.transform(all_entity_strings)
        self.all_entity_ids = all_entity_ids  
    
    @staticmethod
    def init_from_pretrained(pretrained_entity_kb_path: str) -> EntityRetrieval:
        # Error here, cannot define EntityRetrieval 

As I want to wrap it in class and do not want to use a separated function. I guess it’s more professional. So I create a static method. init_from_pretrained but it gave me an error which is not define Entity Retrieval. And the pretrained_entity_kb is binary file which is saved by pickle lib

So in best practice, how to do it?