“model_name”: “llama_3_1”, “model_type”: “transformer”, “num_layers”: 12, “hidden_size”: 768, “num_heads”: 12, “vocab_size”: 32000 ...