{ "architectures": [ "GPTForCausalLM" ], "auto_map": { "AutoConfig": "config.GPTConfig", "AutoModelForCausalLM": "model.GPTForCausalLM" }, "block_size": 512, "dtype": "float32", "head_dim": 64, "hidden_size": 192, "intermediate_size": 480, "labels_are_shifted": true, "max_position_embeddings": 512, "model_type": "gpt", "num_attention_heads": 3, "num_hidden_layers": 7, "num_key_value_heads": 1, "rms_norm_eps": 1e-06, "rope_theta": 5000.0, "transformers_version": "4.57.6", "vocab_size": 4096, "xsa_projection": true }