{ "architectures": [ "LlamaForCausalLM" ], "hidden_size": 4096, "num_hidden_layers": 32, "num_attention_heads": 32, "intermediate_size": 16384, "vocab_size": 50257, "max_position_embeddings": 2048, "type_vocab_size": 2, "initializer_range": 0.02 }