{
    "model_type": "clip",
    "params": {
        "text_encoder_param": "jina_clip_v2_text.ncnn.param",
        "text_encoder_bin": "jina_clip_v2_text.ncnn.bin",
        "vision_encoder_param": "jina_clip_v2_vision.ncnn.param",
        "vision_encoder_bin": "jina_clip_v2_vision.ncnn.bin"
    },
    "tokenizer": {
        "type": "unigram",
        "vocab_file": "vocab.txt",
        "bos": "<s>",
        "eos": "</s>",
        "pad": "<pad>",
        "unk": "<unk>",
        "cls": "<s>",
        "sep": "</s>",
        "mask": "<mask>"
    },
    "setting": {
        "text_embed_dim": 1024,
        "vision_embed_dim": 1024,
        "rope": {
            "type": "RoPE_full",
            "rope_head_dim": 64,
            "rope_theta": 1000000.0
        },
        "image_size": 512,
        "image_mean": [0.485, 0.456, 0.406],
        "image_std": [0.229, 0.224, 0.225],
        "max_seq_len": 77
    }
}