config.json

{
  "architectures": [
    "VitDet3D"
  ],
  "attention_probs_dropout_prob": 0.0,
  "encoder_stride": 16,
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.0,
  "hidden_size": 256,
  "id2label": {
    "0": "LABEL_0"
  },
  "image_size": [
    40,
    128,
    128
  ],
  "initializer_range": 0.02,
  "intermediate_size": 1024,
  "label2id": {
    "LABEL_0": 0
  },
  "layer_norm_eps": 1e-12,
  "model_type": "vit",
  "num_attention_heads": 8,
  "num_channels": 1,
  "num_hidden_layers": 6,
  "patch_size": [
    4,
    16,
    16
  ],
  "qkv_bias": true,
  "torch_dtype": "float32",
  "transformers_version": "4.34.0"
}