config_v5.json

{
  "model_type": "custom_clip_with_registers",

  "processor_class": "CLIPProcessor",
  "tokenizer_class": "CLIPTokenizerFast",

  "architectures": ["CustomCLIPModel"],
  "auto_map": {
    "AutoConfig": "modeling_custom_clip.CustomCLIPConfig", 
    "AutoModel": "modeling_custom_clip.CustomCLIPModel"
  },
  "vision_config": {
    "hidden_size": 1024,
    "num_hidden_layers": 24,
    "num_attention_heads": 16,
    "image_size": 336,
    "patch_size": 14
  },
  "text_config": {
    "vocab_size": 49408,
    "hidden_size": 768,
    "num_hidden_layers": 12,
    "max_position_embeddings": 77
  },
  "num_register_tokens": 3,
  "neuron_dict": {
    "11": [
        3784,
        987,
        3661,
        1967,
        9,
        1100,
        2555
    ],
    "9": [
        815,
        4078,
        3618,
        2693,
        3973,
        1983,
        1744,
        1157
    ],
    "8": [
        745,
        2585,
        3249
    ],
    "10": [
        357,
        1654,
        3940,
        2319,
        2560
    ],
    "7": [
        3228,
        2550
    ]
},
  "projection_dim": 768,      
  "torch_dtype": "float32",
  "transformers_version": "4.21.0",
  "comment": "11 layer부터 시도하는 cfg, not yet"
}