{ "architectures": [ "Qwen3ForCausalLM" ], "model_type": "qwen3", "hidden_size": 512, "intermediate_size": 1344, "num_hidden_layers": 12, "num_attention_heads": 8, "num_key_value_heads": 2, "head_dim": 64, "vocab_size": 32000, "max_position_embeddings": 1024, "rms_norm_eps": 1e-05, "rope_theta": 10000.0, "tie_word_embeddings": true, "hidden_act": "silu", "attention_bias": false, "attention_dropout": 0.0, "use_sliding_window": false, "sliding_window": null, "max_window_layers": 12, "initializer_range": 0.02, "torch_dtype": "float32", "bos_token_id": 0, "eos_token_id": 6, "use_cache": true }