{ "architectures": ["Qwen3ForCausalLM"], "model_type": "qwen3", "vocab_size": 4000, "hidden_size": 192, "intermediate_size": 512, "num_hidden_layers": 16, "num_attention_heads": 3, "num_key_value_heads": 1, "head_dim": 64, "max_position_embeddings": 512, "rms_norm_eps": 1e-06, "rope_theta": 10000.0, "hidden_act": "silu", "tie_word_embeddings": true, "torch_dtype": "float16", "attention_bias": false, "attention_dropout": 0.0, "mlp_bias": false, "sliding_window": null, "use_sliding_window": false, "bos_token_id": 1, "eos_token_id": 5, "pad_token_id": 3 }