{ "architectures": [ "LlamaForCausalLM" ], "model_type": "llama", "vocab_size": 8192, "hidden_size": 192, "intermediate_size": 512, "num_hidden_layers": 9, "num_attention_heads": 6, "num_key_value_heads": 2, "head_dim": 32, "hidden_act": "silu", "max_position_embeddings": 2048, "rope_theta": 10000.0, "rope_scaling": null, "rms_norm_eps": 1e-05, "initializer_range": 0.02, "attention_bias": false, "attention_dropout": 0.0, "mlp_bias": false, "tie_word_embeddings": true, "bos_token_id": 0, "eos_token_id": 0, "pad_token_id": 1, "use_cache": true, "torch_dtype": "float32" }