config.json

{
  "architectures": [
    "M3DCLIP"
  ],
  "auto_map": {
    "AutoConfig": "configuration_m3d_clip.M3DCLIPConfig",
    "AutoModel": "modeling_m3d_clip.M3DCLIP"
  },
  "dropout_rate": 0,
  "gather_loss": true,
  "hidden_size": 768,
  "img_size": [
    32,
    256,
    256
  ],
  "in_channels": 1,
  "language_model_name_or_path": "bert-base-uncased",
  "local_loss": false,
  "max_text_len": 128,
  "mlp_dim": 3072,
  "model_type": "m3d_clip",
  "num_heads": 12,
  "num_layers": 12,
  "patch_size": [
    4,
    16,
    16
  ],
  "pos_embed": "perceptron",
  "spatial_dims": 3,
  "torch_dtype": "float32",
  "transformers_version": "4.40.1",
  "vocab_size": 30522
}