{ | |
"embed_dim": 512, | |
"vision_cfg": { | |
"image_size": 224, | |
"layers": [ | |
3, | |
4, | |
23, | |
3 | |
], | |
"width": 64, | |
"patch_size": null | |
}, | |
"text_cfg": { | |
"context_length": 77, | |
"vocab_size": 49408, | |
"width": 512, | |
"heads": 8, | |
"layers": 12 | |
} | |
} |
{ | |
"embed_dim": 512, | |
"vision_cfg": { | |
"image_size": 224, | |
"layers": [ | |
3, | |
4, | |
23, | |
3 | |
], | |
"width": 64, | |
"patch_size": null | |
}, | |
"text_cfg": { | |
"context_length": 77, | |
"vocab_size": 49408, | |
"width": 512, | |
"heads": 8, | |
"layers": 12 | |
} | |
} |