config.json

{
  "architectures": [
    "YourCaptionModel"
  ],
  "bos_token_id": 1,
  "dropout_prob": 0.5,
  "embed_size": 256,
  "eos_token_id": 2,
  "hidden_size": 256,
  "image_feature_dim": 256,
  "max_caption_len": 20,
  "model_type": "your_caption_model",
  "num_layers_lstm": 1,
  "pad_token_id": 0,
  "sos_token_id": 1,
  "torch_dtype": "float32",
  "transformers_version": "4.52.2",
  "vocab_size": 2892
}