{ "architectures": [ "MllamaTextModel" ], "bos_token_id": 128000, "cross_attention_layers": [ 1 ], "dropout": 0.0, "dtype": "float32", "eos_token_id": 128001, "hidden_act": "gelu", "hidden_size": 32, "initializer_range": 0.02, "intermediate_size": 37, "max_position_embeddings": 512, "model_type": "mllama_text_model", "num_attention_heads": 4, "num_hidden_layers": 2, "num_key_value_heads": 4, "pad_token_id": 128004, "rms_norm_eps": 1e-05, "rope_parameters": { "rope_theta": 500000.0, "rope_type": "default" }, "tie_word_embeddings": false, "transformers_version": "5.16.0.dev0", "use_cache": true, "vocab_size": 128257 }