train/models/vit/base-32-224_config.json

{
  "emb_size": 768,
  "feedforward_size": 3072,
  "hidden_size": 768,
  "hidden_act": "gelu",
  "heads_num": 12,
  "layers_num": 12,
  "dropout": 0.1,
  "max_seq_length": 50,
  "data_processor": "vit",
  "embedding": ["patch", "pos"],
  "remove_embedding_layernorm": true,
  "encoder": "transformer",
  "mask": "fully_visible",
  "layernorm_positioning": "pre",
  "target": ["cls"],
  "image_height": 224,
  "image_width": 224,
  "patch_size": 32
}