XaviXva/Video-LLaVA
0
1o
2 FGe�3 � @ sP d dl Z ddlmZ ddlmZmZ ddlmZ d dlm Z dd� Z4d d5� Z dS ) � N� )�CLIPVisionTower)�LanguageBindImageTower�LanguageBindVideoTower)�MAEVisionTower)� CLIPModelc K s� t | dt | dd ��}tj�|�}|s |�d�s |�d�r&t|fd| i|��S |�d�r6t|f| dd�|��S d |v rYtd6� td7� td8� td9� td10� t |f| dd�|��S t11d |� ���) N�mm_image_tower� image_tower�openaiZlaion�argsZLanguageBind_Image� ./cache_dir�r � cache_dirZmaeZmaemaemaemaemaemaemaemaezUnknown image tower: ) �getattr�os�path�exists�12startswithr �endswithr �printr �13ValueError)Zimage_tower_cfg�kwargsr Zis_absolute_path_exists� r �J/scc_cephfs/yy/lb/LLaVA-Video-YY/llava/model/multimodal_encoder/builder.py�build_image_tower s 14r c K sB t | dt | dd ��}|�d�rt|f| dd�|��S td|� ���)N�mm_video_tower� video_towerZLanguageBind_Videor r
zUnknown video tower: )r r r r )Zvideo_tower_cfgr r r r r �build_video_tower s 15r ) r � clip_encoderr Z languagebindr r Z mae_encoderr � transformersr r r r r r r �<module> s