{ "architectures": ["GemmaForCausalLM"], "model_type": "gemma", "hidden_size": 1024, "num_attention_heads": 16, "num_hidden_layers": 24, "vocab_size": 262144 }