{"model_type": "encoder-decoder", "vocab_size": 50000, "d_model": 256, "nhead": 8, "num_encoder_layers": 2, "num_decoder_layers": 2, "dim_feedforward": 2048, "architectures": ["EncoderDecoderModel"], "attention_probs_dropout_prob": 0.1, "hidden_dropout_prob": 0.1, "model_specific_params": {"model_type": "encoder-decoder", "layer_norm_eps": 1e-05, "activation_function": "gelu", "hidden_size": 256, "num_attention_heads": 8, "num_hidden_layers": 2, "intermediate_size": 2048, "max_position_embeddings": 512, "vocab_size": 50000, "pad_token_id": 0, "bos_token_id": 1, "eos_token_id": 2}}