config.json

{
  "_name_or_path": "Mitsua/mitsua-japanese-clip-vit-b-16",
  "architectures": [
    "MitsuaJapaneseCLIPModel"
  ],
  "auto_map": {
    "AutoConfig": "configuration_mitsua_japanese_clip.MitsuaJapaneseCLIPConfig",
    "AutoModel": "modeling_mitsua_japanese_clip.MitsuaJapaneseCLIPModel"
  },
  "initializer_factor": 1.0,
  "logit_scale_init_value": 2.6592,
  "model_type": "mitsua_japanese_clip",
  "projection_dim": 768,
  "text_config": {
    "bos_token_id": -1,
    "eos_token_id": 1,
    "hidden_act": "gelu",
    "layer_norm_eps": 1e-05,
    "model_type": "siglip_text_model",
    "vocab_size": 64000
  },
  "torch_dtype": "float32",
  "transformers_version": "4.44.2",
  "vision_config": {
    "hidden_act": "gelu",
    "layer_norm_eps": 1e-06,
    "model_type": "clip_vision_model",
    "patch_size": 16,
    "projection_dim": 768
  }
}