{ "architectures": [ "Qwen3ForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "dtype": "float32", "eos_token_id": 151643, "head_dim": 128, "hidden_act": "silu", "hidden_size": 2560, "initializer_range": 0.02, "intermediate_size": 9728, "layer_types": [ "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention", "full_attention" ], "max_position_embeddings": 32768, "max_window_layers": 28, "model_type": "qwen3", "num_attention_heads": 32, "num_hidden_layers": 36, "num_key_value_heads": 8, "pad_token_id": 151643, "quantization_config": { "include_input_output_embeddings": true, "modules_to_not_convert": [ "model.layers.0.self_attn.q_norm", "model.layers.0.self_attn.k_norm", "model.layers.0.input_layernorm", "model.layers.0.post_attention_layernorm", "model.layers.1.self_attn.q_norm", "model.layers.1.self_attn.k_norm", "model.layers.1.input_layernorm", "model.layers.1.post_attention_layernorm", "model.layers.2.self_attn.q_norm", "model.layers.2.self_attn.k_norm", "model.layers.2.input_layernorm", "model.layers.2.post_attention_layernorm", "model.layers.3.self_attn.q_norm", "model.layers.3.self_attn.k_norm", "model.layers.3.input_layernorm", "model.layers.3.post_attention_layernorm", "model.layers.4.self_attn.q_norm", "model.layers.4.self_attn.k_norm", "model.layers.4.input_layernorm", "model.layers.4.post_attention_layernorm", "model.layers.5.self_attn.q_norm", "model.layers.5.self_attn.k_norm", "model.layers.5.input_layernorm", "model.layers.5.post_attention_layernorm", "model.layers.6.self_attn.q_norm", "model.layers.6.self_attn.k_norm", "model.layers.6.input_layernorm", "model.layers.6.post_attention_layernorm", "model.layers.7.self_attn.q_norm", "model.layers.7.self_attn.k_norm", "model.layers.7.input_layernorm", "model.layers.7.post_attention_layernorm", "model.layers.8.self_attn.q_norm", "model.layers.8.self_attn.k_norm", "model.layers.8.input_layernorm", "model.layers.8.post_attention_layernorm", "model.layers.9.self_attn.q_norm", "model.layers.9.self_attn.k_norm", "model.layers.9.input_layernorm", "model.layers.9.post_attention_layernorm", "model.layers.10.self_attn.q_norm", "model.layers.10.self_attn.k_norm", "model.layers.10.input_layernorm", "model.layers.10.post_attention_layernorm", "model.layers.11.self_attn.q_norm", "model.layers.11.self_attn.k_norm", "model.layers.11.input_layernorm", "model.layers.11.post_attention_layernorm", "model.layers.12.self_attn.q_norm", "model.layers.12.self_attn.k_norm", "model.layers.12.input_layernorm", "model.layers.12.post_attention_layernorm", "model.layers.13.self_attn.q_norm", "model.layers.13.self_attn.k_norm", "model.layers.13.input_layernorm", "model.layers.13.post_attention_layernorm", "model.layers.14.self_attn.q_norm", "model.layers.14.self_attn.k_norm", "model.layers.14.input_layernorm", "model.layers.14.post_attention_layernorm", "model.layers.15.self_attn.q_norm", "model.layers.15.self_attn.k_norm", "model.layers.15.input_layernorm", "model.layers.15.post_attention_layernorm", "model.layers.16.self_attn.q_norm", "model.layers.16.self_attn.k_norm", "model.layers.16.input_layernorm", "model.layers.16.post_attention_layernorm", "model.layers.17.self_attn.q_norm", "model.layers.17.self_attn.k_norm", "model.layers.17.input_layernorm", "model.layers.17.post_attention_layernorm", "model.layers.18.self_attn.q_norm", "model.layers.18.self_attn.k_norm", "model.layers.18.input_layernorm", "model.layers.18.post_attention_layernorm", "model.layers.19.self_attn.q_norm", "model.layers.19.self_attn.k_norm", "model.layers.19.input_layernorm", "model.layers.19.post_attention_layernorm", "model.layers.20.self_attn.q_norm", "model.layers.20.self_attn.k_norm", "model.layers.20.input_layernorm", "model.layers.20.post_attention_layernorm", "model.layers.21.self_attn.q_norm", "model.layers.21.self_attn.k_norm", "model.layers.21.input_layernorm", "model.layers.21.post_attention_layernorm", "model.layers.22.self_attn.q_norm", "model.layers.22.self_attn.k_norm", "model.layers.22.input_layernorm", "model.layers.22.post_attention_layernorm", "model.layers.23.self_attn.q_norm", "model.layers.23.self_attn.k_norm", "model.layers.23.input_layernorm", "model.layers.23.post_attention_layernorm", "model.layers.24.self_attn.q_norm", "model.layers.24.self_attn.k_norm", "model.layers.24.input_layernorm", "model.layers.24.post_attention_layernorm", "model.layers.25.self_attn.q_norm", "model.layers.25.self_attn.k_norm", "model.layers.25.input_layernorm", "model.layers.25.post_attention_layernorm", "model.layers.26.self_attn.q_norm", "model.layers.26.self_attn.k_norm", "model.layers.26.input_layernorm", "model.layers.26.post_attention_layernorm", "model.layers.27.self_attn.q_norm", "model.layers.27.self_attn.k_norm", "model.layers.27.input_layernorm", "model.layers.27.post_attention_layernorm", "model.layers.28.self_attn.q_norm", "model.layers.28.self_attn.k_norm", "model.layers.28.input_layernorm", "model.layers.28.post_attention_layernorm", "model.layers.29.self_attn.q_norm", "model.layers.29.self_attn.k_norm", "model.layers.29.input_layernorm", "model.layers.29.post_attention_layernorm", "model.layers.30.self_attn.q_norm", "model.layers.30.self_attn.k_norm", "model.layers.30.input_layernorm", "model.layers.30.post_attention_layernorm", "model.layers.31.self_attn.q_norm", "model.layers.31.self_attn.k_norm", "model.layers.31.input_layernorm", "model.layers.31.post_attention_layernorm", "model.layers.32.self_attn.q_norm", "model.layers.32.self_attn.k_norm", "model.layers.32.input_layernorm", "model.layers.32.post_attention_layernorm", "model.layers.33.self_attn.q_norm", "model.layers.33.self_attn.k_norm", "model.layers.33.input_layernorm", "model.layers.33.post_attention_layernorm", "model.layers.34.self_attn.q_norm", "model.layers.34.self_attn.k_norm", "model.layers.34.input_layernorm", "model.layers.34.post_attention_layernorm", "model.layers.35.self_attn.q_norm", "model.layers.35.self_attn.k_norm", "model.layers.35.input_layernorm", "model.layers.35.post_attention_layernorm", "model.norm", "lm_head" ], "quant_method": "torchao", "quant_type": { "default": { "_data": { "module_fqn_to_config": { "_default": { "_data": { "activation_quantization": "int8_asym_per_token", "b": 2, "granularity": { "_data": { "axis": 0 }, "_type": "PerAxis", "_version": 1 }, "layout": { "_data": {}, "_type": "QDQLayout", "_version": 1 }, "quant_max": 1.5, "quant_min": -1.5, "scale_dtype": null }, "_type": "StretchedIntxWeightConfig", "_version": 2 }, "model.embed_tokens": { "_data": { "act_mapping_type": { "_data": "ASYMMETRIC", "_type": "MappingType" }, "intx_packing_format": { "_data": "UNPACKED_TO_INT8", "_type": "IntxPackingFormat" }, "layout": { "_data": {}, "_type": "QDQLayout", "_version": 1 }, "weight_dtype": { "_data": "int4", "_type": "torch.dtype" }, "weight_granularity": { "_data": { "axis": 0 }, "_type": "PerAxis", "_version": 1 }, "weight_mapping_type": { "_data": "SYMMETRIC", "_type": "MappingType" }, "weight_scale_dtype": null }, "_type": "Int8DynamicActivationIntxWeightConfig", "_version": 2 } } }, "_type": "ModuleFqnToConfig", "_version": 1 } }, "quant_type_kwargs": {}, "untie_embedding_weights": false }, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 1000000.0, "sliding_window": null, "tie_word_embeddings": true, "transformers_version": "4.56.1", "use_cache": false, "use_sliding_window": false, "vocab_size": 151936 }