search

THUDM / CogVLM

a state-of-the-art-level open visual language model | 多模态预训练模型
Apache License 2.0
5.92k stars 407 forks source link

CogAgent 视觉预训练模型 EVA2-CLIP-L #474

Open hzhiyuan opened 4 months ago

hzhiyuan commented 4 months ago

想用 CogAgent 的视觉预训练模型，看到 https://github.com/THUDM/SwissArmyTransformer/blob/main/sat/resources/urls.py 里面只有一个名为 eva02_L_pt_m38m_p14 的模型，请问 eva02_L_pt_m38m_p14 就是 CogAgent 的视觉预训练模型吗 @1049451037

1049451037 commented 4 months ago

eva-clip-4b-14-x-drop-last-layer 是CogAgent的预训练初始化，对应 sat/model/official/eva_clip_model.py

eva02_L_pt_m38m_p14 是另一个vit模型，对应 sat/model/official/eva2_model.py