{ "architectures": [ "LlamaForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 1, "eos_token_id": 7, "hidden_act": "silu", "hidden_size": 7168, "initializer_range": 0.02, "intermediate_size": 20480, "max_position_embeddings": 32768, "mlp_bias": false, "model_type": "llama", "num_attention_heads": 56, "num_hidden_layers": 60, "num_key_value_heads": 8, "pretraining_tp": 1, "quantization": { "group_size": 64, "bits": "4" }, "rms_norm_eps": 1e-06, "rope_scaling": null, "rope_theta": 5000000, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "transformers_version": "4.41.1", "use_cache": false, "vocab_size": 64000 }