{ "architectures": [ "MiniMaxM2ForCausalLM" ], "attention_dropout": 0.0, "attn_type_list": [ 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 ], "auto_map": { "AutoConfig": "configuration_minimax_m2.MiniMaxM2Config", "AutoModelForCausalLM": "modeling_minimax_m2.MiniMaxM2ForCausalLM" }, "bos_token_id": 1, "dtype": "bfloat16", "eos_token_id": 2, "head_dim": 128, "hidden_act": "silu", "hidden_size": 3072, "initializer_range": 0.02, "intermediate_size": 1536, "max_position_embeddings": 196608, "model_type": "minimax_m2", "mtp_transformer_layers": 1, "num_attention_heads": 48, "num_experts_per_tok": 8, "num_hidden_layers": 62, "num_key_value_heads": 8, "num_local_experts": 256, "num_mtp_modules": 3, "output_router_logits": false, "partial_rotary_factor": 0.5, "qk_norm_type": "per_layer", "quantization_config": { "algo_config": null, "exclude": [ "model.layers.0.self_attn.q_proj", "model.layers.0.self_attn.k_proj", "model.layers.0.self_attn.v_proj", "model.layers.0.self_attn.o_proj", "model.layers.0.block_sparse_moe.gate", "model.layers.1.self_attn.q_proj", "model.layers.1.self_attn.k_proj", "model.layers.1.self_attn.v_proj", "model.layers.1.self_attn.o_proj", "model.layers.1.block_sparse_moe.gate", "model.layers.2.self_attn.q_proj", "model.layers.2.self_attn.k_proj", "model.layers.2.self_attn.v_proj", "model.layers.2.self_attn.o_proj", "model.layers.2.block_sparse_moe.gate", "model.layers.3.self_attn.q_proj", "model.layers.3.self_attn.k_proj", "model.layers.3.self_attn.v_proj", "model.layers.3.self_attn.o_proj", "model.layers.3.block_sparse_moe.gate", "model.layers.4.self_attn.q_proj", "model.layers.4.self_attn.k_proj", "model.layers.4.self_attn.v_proj", "model.layers.4.self_attn.o_proj", "model.layers.4.block_sparse_moe.gate", "model.layers.5.self_attn.q_proj", "model.layers.5.self_attn.k_proj", "model.layers.5.self_attn.v_proj", "model.layers.5.self_attn.o_proj", "model.layers.5.block_sparse_moe.gate", "model.layers.6.self_attn.q_proj", "model.layers.6.self_attn.k_proj", "model.layers.6.self_attn.v_proj", "model.layers.6.self_attn.o_proj", "model.layers.6.block_sparse_moe.gate", "model.layers.7.self_attn.q_proj", "model.layers.7.self_attn.k_proj", "model.layers.7.self_attn.v_proj", "model.layers.7.self_attn.o_proj", "model.layers.7.block_sparse_moe.gate", "model.layers.8.self_attn.q_proj", "model.layers.8.self_attn.k_proj", "model.layers.8.self_attn.v_proj", "model.layers.8.self_attn.o_proj", "model.layers.8.block_sparse_moe.gate", "model.layers.9.self_attn.q_proj", "model.layers.9.self_attn.k_proj", "model.layers.9.self_attn.v_proj", "model.layers.9.self_attn.o_proj", "model.layers.9.block_sparse_moe.gate", "model.layers.10.self_attn.q_proj", "model.layers.10.self_attn.k_proj", "model.layers.10.self_attn.v_proj", "model.layers.10.self_attn.o_proj", "model.layers.10.block_sparse_moe.gate", "model.layers.11.self_attn.q_proj", "model.layers.11.self_attn.k_proj", "model.layers.11.self_attn.v_proj", "model.layers.11.self_attn.o_proj", "model.layers.11.block_sparse_moe.gate", "model.layers.12.self_attn.q_proj", "model.layers.12.self_attn.k_proj", "model.layers.12.self_attn.v_proj", "model.layers.12.self_attn.o_proj", "model.layers.12.block_sparse_moe.gate", "model.layers.13.self_attn.q_proj", "model.layers.13.self_attn.k_proj", "model.layers.13.self_attn.v_proj", "model.layers.13.self_attn.o_proj", "model.layers.13.block_sparse_moe.gate", "model.layers.14.self_attn.q_proj", "model.layers.14.self_attn.k_proj", "model.layers.14.self_attn.v_proj", "model.layers.14.self_attn.o_proj", "model.layers.14.block_sparse_moe.gate", "model.layers.15.self_attn.q_proj", "model.layers.15.self_attn.k_proj", "model.layers.15.self_attn.v_proj", "model.layers.15.self_attn.o_proj", "model.layers.15.block_sparse_moe.gate", "model.layers.16.self_attn.q_proj", "model.layers.16.self_attn.k_proj", "model.layers.16.self_attn.v_proj", "model.layers.16.self_attn.o_proj", "model.layers.16.block_sparse_moe.gate", "model.layers.17.self_attn.q_proj", "model.layers.17.self_attn.k_proj", "model.layers.17.self_attn.v_proj", "model.layers.17.self_attn.o_proj", "model.layers.17.block_sparse_moe.gate", "model.layers.18.self_attn.q_proj", "model.layers.18.self_attn.k_proj", "model.layers.18.self_attn.v_proj", "model.layers.18.self_attn.o_proj", "model.layers.18.block_sparse_moe.gate", "model.layers.19.self_attn.q_proj", "model.layers.19.self_attn.k_proj", "model.layers.19.self_attn.v_proj", "model.layers.19.self_attn.o_proj", "model.layers.19.block_sparse_moe.gate", "model.layers.20.self_attn.q_proj", "model.layers.20.self_attn.k_proj", "model.layers.20.self_attn.v_proj", "model.layers.20.self_attn.o_proj", "model.layers.20.block_sparse_moe.gate", "model.layers.21.self_attn.q_proj", "model.layers.21.self_attn.k_proj", "model.layers.21.self_attn.v_proj", "model.layers.21.self_attn.o_proj", "model.layers.21.block_sparse_moe.gate", "model.layers.22.self_attn.q_proj", "model.layers.22.self_attn.k_proj", "model.layers.22.self_attn.v_proj", "model.layers.22.self_attn.o_proj", "model.layers.22.block_sparse_moe.gate", "model.layers.23.self_attn.q_proj", "model.layers.23.self_attn.k_proj", "model.layers.23.self_attn.v_proj", "model.layers.23.self_attn.o_proj", "model.layers.23.block_sparse_moe.gate", "model.layers.24.self_attn.q_proj", "model.layers.24.self_attn.k_proj", "model.layers.24.self_attn.v_proj", "model.layers.24.self_attn.o_proj", "model.layers.24.block_sparse_moe.gate", "model.layers.25.self_attn.q_proj", "model.layers.25.self_attn.k_proj", "model.layers.25.self_attn.v_proj", "model.layers.25.self_attn.o_proj", "model.layers.25.block_sparse_moe.gate", "model.layers.26.self_attn.q_proj", "model.layers.26.self_attn.k_proj", "model.layers.26.self_attn.v_proj", "model.layers.26.self_attn.o_proj", "model.layers.26.block_sparse_moe.gate", "model.layers.27.self_attn.q_proj", "model.layers.27.self_attn.k_proj", "model.layers.27.self_attn.v_proj", "model.layers.27.self_attn.o_proj", "model.layers.27.block_sparse_moe.gate", "model.layers.28.self_attn.q_proj", "model.layers.28.self_attn.k_proj", "model.layers.28.self_attn.v_proj", "model.layers.28.self_attn.o_proj", "model.layers.28.block_sparse_moe.gate", "model.layers.29.self_attn.q_proj", "model.layers.29.self_attn.k_proj", "model.layers.29.self_attn.v_proj", "model.layers.29.self_attn.o_proj", "model.layers.29.block_sparse_moe.gate", "model.layers.30.self_attn.q_proj", "model.layers.30.self_attn.k_proj", "model.layers.30.self_attn.v_proj", "model.layers.30.self_attn.o_proj", "model.layers.30.block_sparse_moe.gate", "model.layers.31.self_attn.q_proj", "model.layers.31.self_attn.k_proj", "model.layers.31.self_attn.v_proj", "model.layers.31.self_attn.o_proj", "model.layers.31.block_sparse_moe.gate", "model.layers.32.self_attn.q_proj", "model.layers.32.self_attn.k_proj", "model.layers.32.self_attn.v_proj", "model.layers.32.self_attn.o_proj", "model.layers.32.block_sparse_moe.gate", "model.layers.33.self_attn.q_proj", "model.layers.33.self_attn.k_proj", "model.layers.33.self_attn.v_proj", "model.layers.33.self_attn.o_proj", "model.layers.33.block_sparse_moe.gate", "model.layers.34.self_attn.q_proj", "model.layers.34.self_attn.k_proj", "model.layers.34.self_attn.v_proj", "model.layers.34.self_attn.o_proj", "model.layers.34.block_sparse_moe.gate", "model.layers.35.self_attn.q_proj", "model.layers.35.self_attn.k_proj", "model.layers.35.self_attn.v_proj", "model.layers.35.self_attn.o_proj", "model.layers.35.block_sparse_moe.gate", "model.layers.36.self_attn.q_proj", "model.layers.36.self_attn.k_proj", "model.layers.36.self_attn.v_proj", "model.layers.36.self_attn.o_proj", "model.layers.36.block_sparse_moe.gate", "model.layers.37.self_attn.q_proj", "model.layers.37.self_attn.k_proj", "model.layers.37.self_attn.v_proj", "model.layers.37.self_attn.o_proj", "model.layers.37.block_sparse_moe.gate", "model.layers.38.self_attn.q_proj", "model.layers.38.self_attn.k_proj", "model.layers.38.self_attn.v_proj", "model.layers.38.self_attn.o_proj", "model.layers.38.block_sparse_moe.gate", "model.layers.39.self_attn.q_proj", "model.layers.39.self_attn.k_proj", "model.layers.39.self_attn.v_proj", "model.layers.39.self_attn.o_proj", "model.layers.39.block_sparse_moe.gate", "model.layers.40.self_attn.q_proj", "model.layers.40.self_attn.k_proj", "model.layers.40.self_attn.v_proj", "model.layers.40.self_attn.o_proj", "model.layers.40.block_sparse_moe.gate", "model.layers.41.self_attn.q_proj", "model.layers.41.self_attn.k_proj", "model.layers.41.self_attn.v_proj", "model.layers.41.self_attn.o_proj", "model.layers.41.block_sparse_moe.gate", "model.layers.42.self_attn.q_proj", "model.layers.42.self_attn.k_proj", "model.layers.42.self_attn.v_proj", "model.layers.42.self_attn.o_proj", "model.layers.42.block_sparse_moe.gate", "model.layers.43.self_attn.q_proj", "model.layers.43.self_attn.k_proj", "model.layers.43.self_attn.v_proj", "model.layers.43.self_attn.o_proj", "model.layers.43.block_sparse_moe.gate", "model.layers.44.self_attn.q_proj", "model.layers.44.self_attn.k_proj", "model.layers.44.self_attn.v_proj", "model.layers.44.self_attn.o_proj", "model.layers.44.block_sparse_moe.gate", "model.layers.45.self_attn.q_proj", "model.layers.45.self_attn.k_proj", "model.layers.45.self_attn.v_proj", "model.layers.45.self_attn.o_proj", "model.layers.45.block_sparse_moe.gate", "model.layers.46.self_attn.q_proj", "model.layers.46.self_attn.k_proj", "model.layers.46.self_attn.v_proj", "model.layers.46.self_attn.o_proj", "model.layers.46.block_sparse_moe.gate", "model.layers.47.self_attn.q_proj", "model.layers.47.self_attn.k_proj", "model.layers.47.self_attn.v_proj", "model.layers.47.self_attn.o_proj", "model.layers.47.block_sparse_moe.gate", "model.layers.48.self_attn.q_proj", "model.layers.48.self_attn.k_proj", "model.layers.48.self_attn.v_proj", "model.layers.48.self_attn.o_proj", "model.layers.48.block_sparse_moe.gate", "model.layers.49.self_attn.q_proj", "model.layers.49.self_attn.k_proj", "model.layers.49.self_attn.v_proj", "model.layers.49.self_attn.o_proj", "model.layers.49.block_sparse_moe.gate", "model.layers.50.self_attn.q_proj", "model.layers.50.self_attn.k_proj", "model.layers.50.self_attn.v_proj", "model.layers.50.self_attn.o_proj", "model.layers.50.block_sparse_moe.gate", "model.layers.51.self_attn.q_proj", "model.layers.51.self_attn.k_proj", "model.layers.51.self_attn.v_proj", "model.layers.51.self_attn.o_proj", "model.layers.51.block_sparse_moe.gate", "model.layers.52.self_attn.q_proj", "model.layers.52.self_attn.k_proj", "model.layers.52.self_attn.v_proj", "model.layers.52.self_attn.o_proj", "model.layers.52.block_sparse_moe.gate", "model.layers.53.self_attn.q_proj", "model.layers.53.self_attn.k_proj", "model.layers.53.self_attn.v_proj", "model.layers.53.self_attn.o_proj", "model.layers.53.block_sparse_moe.gate", "model.layers.54.self_attn.q_proj", "model.layers.54.self_attn.k_proj", "model.layers.54.self_attn.v_proj", "model.layers.54.self_attn.o_proj", "model.layers.54.block_sparse_moe.gate", "model.layers.55.self_attn.q_proj", "model.layers.55.self_attn.k_proj", "model.layers.55.self_attn.v_proj", "model.layers.55.self_attn.o_proj", "model.layers.55.block_sparse_moe.gate", "model.layers.56.self_attn.q_proj", "model.layers.56.self_attn.k_proj", "model.layers.56.self_attn.v_proj", "model.layers.56.self_attn.o_proj", "model.layers.56.block_sparse_moe.gate", "model.layers.57.self_attn.q_proj", "model.layers.57.self_attn.k_proj", "model.layers.57.self_attn.v_proj", "model.layers.57.self_attn.o_proj", "model.layers.57.block_sparse_moe.gate", "model.layers.58.self_attn.q_proj", "model.layers.58.self_attn.k_proj", "model.layers.58.self_attn.v_proj", "model.layers.58.self_attn.o_proj", "model.layers.58.block_sparse_moe.gate", "model.layers.59.self_attn.q_proj", "model.layers.59.self_attn.k_proj", "model.layers.59.self_attn.v_proj", "model.layers.59.self_attn.o_proj", "model.layers.59.block_sparse_moe.gate", "model.layers.60.self_attn.q_proj", "model.layers.60.self_attn.k_proj", "model.layers.60.self_attn.v_proj", "model.layers.60.self_attn.o_proj", "model.layers.60.block_sparse_moe.gate", "model.layers.61.self_attn.q_proj", "model.layers.61.self_attn.k_proj", "model.layers.61.self_attn.v_proj", "model.layers.61.self_attn.o_proj", "model.layers.61.block_sparse_moe.gate", "lm_head" ], "export": { "kv_cache_group": [], "min_kv_scale": 0.0, "pack_method": "reorder", "weight_format": "real_quantized", "weight_merge_groups": null }, "global_quant_config": { "bias": null, "input_tensors": { "ch_axis": -1, "dtype": "fp4", "group_size": 32, "is_dynamic": true, "is_scale_quant": false, "mx_element_dtype": null, "observer_cls": "PerBlockMXObserver", "qscheme": "per_group", "round_method": "half_even", "scale_calculation_mode": "even", "scale_format": "e8m0", "scale_type": "float", "symmetric": null }, "output_tensors": null, "target_device": null, "weight": { "ch_axis": -1, "dtype": "fp4", "group_size": 32, "is_dynamic": false, "is_scale_quant": false, "mx_element_dtype": null, "observer_cls": "PerBlockMXObserver", "qscheme": "per_group", "round_method": "half_even", "scale_calculation_mode": "even", "scale_format": "e8m0", "scale_type": "float", "symmetric": null } }, "kv_cache_post_rope": false, "kv_cache_quant_config": {}, "layer_quant_config": {}, "layer_type_quant_config": {}, "quant_method": "quark", "quant_mode": "eager_mode", "softmax_quant_spec": null, "version": "0.11" }, "rms_norm_eps": 1e-06, "rope_theta": 5000000, "rotary_dim": 64, "router_aux_loss_coef": 0.001, "router_jitter_noise": 0.0, "scoring_func": "sigmoid", "shared_intermediate_size": 0, "sliding_window": null, "tie_word_embeddings": false, "transformers_version": "4.57.1", "use_cache": true, "use_mtp": true, "use_qk_norm": true, "use_routing_bias": true, "vocab_size": 200064 }