{
  "architectures": [
    "CLIPModel"
  ],
  "initializer_factor": 1.0,
  "logit_scale_init_value": 2.6592,
  "model_type": "clip",
  "projection_dim": 512,
  "text_config": {
    "_name_or_path": "/root/.cache/huggingface/hub/models--stabilityai--stable-diffusion-2-1-unclip/snapshots/e99f66a92bdcd1b0fb0d4b6a9b81b3b37d8bea44/text_encoder",
    "architectures": [
      "CLIPTextModel"
    ],
    "bos_token_id": 0,
    "dropout": 0.0,
    "eos_token_id": 2,
    "hidden_act": "gelu",
    "hidden_size": 1024,
    "intermediate_size": 4096,
    "model_type": "clip_text_model",
    "num_attention_heads": 16,
    "num_hidden_layers": 23,
    "torch_dtype": "float16"
  },
  "torch_dtype": "float32",
  "transformers_version": "4.41.1",
  "vision_config": {
    "_name_or_path": "/root/.cache/huggingface/hub/models--stabilityai--stable-diffusion-2-1-unclip/snapshots/e99f66a92bdcd1b0fb0d4b6a9b81b3b37d8bea44/image_encoder",
    "architectures": [
      "CLIPVisionModelWithProjection"
    ],
    "dropout": 0.0,
    "hidden_act": "gelu",
    "hidden_size": 1280,
    "intermediate_size": 5120,
    "model_type": "clip_vision_model",
    "num_attention_heads": 16,
    "num_hidden_layers": 32,
    "patch_size": 14,
    "projection_dim": 1024,
    "torch_dtype": "float16"
  }
}