--- a +++ b/lavis/configs/models/clip/ViT-B-16-plus.json @@ -0,0 +1,16 @@ +{ + "embed_dim": 640, + "vision_cfg": { + "image_size": 224, + "layers": 12, + "width": 896, + "patch_size": 16 + }, + "text_cfg": { + "context_length": 77, + "vocab_size": 49408, + "width": 640, + "heads": 10, + "layers": 12 + } +}