{
  "architectures": [
    "CustomTransformerForCausalLM"
  ],
  "bos_token_id": 2,
  "d_ff": 2048,
  "d_model": 512,
  "dropout": 0.1,
  "dtype": "float32",
  "eos_token_id": 3,
  "max_seq_len": 512,
  "model_type": "custom_transformer",
  "n_heads": 8,
  "n_layers": 6,
  "num_attention_heads": 8,
  "num_hidden_layers": 6,
  "pad_token_id": 0,
  "transformers_version": "4.57.3",
  "vocab_size": 50000
}