{
  "architecture": "CNN(ResNet-50)-Transformer Encoder-Decoder Image Captioning",
  "vocab_size": 8000,
  "d_model": 512,
  "nhead": 8,
  "num_decoder_layers": 6,
  "dim_feedforward": 2048,
  "image_size": 224,
  "max_len": 300,
  "pad_id": 0,
  "unk_id": 1,
  "bos_id": 2,
  "eos_id": 3,
  "encoder_finetune_strategy": "phase1_frozen_then_phase2_layer4_unfrozen_discriminative_lr"
}