{ "architectures": [ "Ernie4_5ForCausalLM" ], "bos_token_id": 1, "dtype": "float32", "eos_token_id": 2, "head_dim": 16, "hidden_act": "gelu", "hidden_size": 32, "initializer_range": 0.02, "intermediate_size": 32, "is_decoder": true, "max_position_embeddings": 512, "model_type": "ernie4_5", "num_attention_heads": 2, "num_hidden_layers": 2, "num_key_value_heads": 2, "pad_token_id": 0, "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_bias": false, "use_cache": true, "vocab_size": 101304 }