def __init__()

in models/vision_language_model.py [0:0]

• 13 lines of code
• 2 McCabe index (conditional complexity)

    def __init__(self, cfg: VLMConfig, load_backbone=True):
        super().__init__()
        self.cfg = cfg
        if load_backbone:
            print("Loading from backbone weights")
            self.vision_encoder = ViT.from_pretrained(cfg)
            self.decoder = LanguageModel.from_pretrained(cfg)
        else:
            self.vision_encoder = ViT(cfg)
            self.decoder = LanguageModel(cfg)
        self.MP = ModalityProjector(cfg)
        self.load_backbone = load_backbone
        self.tokenizer = get_tokenizer(cfg.lm_tokenizer, cfg.vlm_extra_tokens, cfg.lm_chat_template)