{ "architectures": [ "NemotronHForCausalLM" ], "attention_bias": false, "attention_dropout": 0.0, "auto_map": { "AutoConfig": "configuration_nemotron_h.NemotronHConfig", "AutoModelForCausalLM": "modeling_nemotron_h.NemotronHForCausalLM" }, "bos_token_id": 1, "chunk_size": 128, "conv_kernel": 4, "dtype": "bfloat16", "eos_token_id": [ 2, 11 ], "expand": 2, "head_dim": 128, "hidden_dropout": 0.0, "hidden_size": 4096, "hybrid_override_pattern": "MEMEMEM*EMEMEMEM*EMEMEMEM*EMEMEMEMEM*EMEMEMEMEM*EMEMEMEMEM*EMEMEMEMEM*EMEMEMEM*EMEMEMEME", "initializer_range": 0.02, "intermediate_size": 2688, "layer_norm_epsilon": 1e-05, "mamba_head_dim": 64, "mamba_hidden_act": "silu", "mamba_num_heads": 128, "mamba_proj_bias": false, "mamba_ssm_cache_dtype": "float32", "max_position_embeddings": 262144, "mlp_bias": false, "mlp_hidden_act": "relu2", "model_type": "nemotron_h", "moe_intermediate_size": 2688, "moe_latent_size": 1024, "moe_shared_expert_intermediate_size": 5376, "moe_shared_expert_overlap": false, "mtp_hybrid_override_pattern": "*E", "n_group": 1, "n_groups": 8, "n_routed_experts": 512, "n_shared_experts": 1, "norm_eps": 1e-05, "norm_topk_prob": true, "num_attention_heads": 32, "num_experts_per_tok": 22, "num_hidden_layers": 88, "num_key_value_heads": 2, "num_logits_to_keep": 1, "num_nextn_predict_layers": 1, "pad_token_id": 0, "partial_rotary_factor": 1.0, "quantization": { "group_size": 64, "bits": 4, "mode": "affine", "backbone.embeddings": { "bits": 4, "group_size": 64 }, "backbone.layers.0.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.0.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.1.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.1.mixer.switch_mlp.fc2": { "bits": 4, "group_size": 64 }, "backbone.layers.1.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.1.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.1.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.1.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.2.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.2.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.3.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.3.mixer.switch_mlp.fc2": { "bits": 4, "group_size": 64 }, "backbone.layers.3.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.3.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.3.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.3.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.4.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.4.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.5.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.5.mixer.switch_mlp.fc2": { "bits": 4, "group_size": 64 }, "backbone.layers.5.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.5.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.5.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.5.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.6.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.6.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.7.mixer.q_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.7.mixer.k_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.7.mixer.v_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.7.mixer.o_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.8.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.8.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.8.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.8.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.8.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.8.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.9.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.9.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.10.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.10.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.10.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.10.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.10.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.10.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.11.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.11.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.12.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.12.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.12.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.12.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.12.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.12.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.13.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.13.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.14.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.14.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.14.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.14.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.14.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.14.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.15.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.15.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.16.mixer.q_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.16.mixer.k_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.16.mixer.v_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.16.mixer.o_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.17.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.17.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.17.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.17.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.17.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.17.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.18.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.18.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.19.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.19.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.19.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.19.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.19.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.19.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.20.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.20.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.21.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.21.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.21.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.21.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.21.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.21.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.22.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.22.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.23.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.23.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.23.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.23.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.23.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.23.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.24.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.24.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.25.mixer.q_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.25.mixer.k_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.25.mixer.v_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.25.mixer.o_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.26.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.26.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.26.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.26.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.26.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.26.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.27.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.27.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.28.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.28.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.28.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.28.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.28.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.28.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.29.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.29.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.30.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.30.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.30.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.30.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.30.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.30.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.31.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.31.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.32.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.32.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.32.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.32.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.32.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.32.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.33.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.33.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.34.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.34.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.34.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.34.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.34.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.34.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.35.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.35.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.36.mixer.q_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.36.mixer.k_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.36.mixer.v_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.36.mixer.o_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.37.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.37.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.37.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.37.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.37.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.37.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.38.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.38.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.39.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.39.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.39.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.39.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.39.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.39.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.40.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.40.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.41.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.41.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.41.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.41.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.41.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.41.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.42.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.42.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.43.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.43.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.43.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.43.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.43.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.43.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.44.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.44.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.45.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.45.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.45.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.45.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.45.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.45.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.46.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.46.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.47.mixer.q_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.47.mixer.k_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.47.mixer.v_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.47.mixer.o_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.48.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.48.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.48.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.48.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.48.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.48.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.49.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.49.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.50.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.50.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.50.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.50.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.50.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.50.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.51.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.51.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.52.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.52.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.52.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.52.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.52.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.52.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.53.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.53.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.54.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.54.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.54.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.54.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.54.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.54.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.55.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.55.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.56.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.56.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.56.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.56.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.56.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.56.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.57.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.57.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.58.mixer.q_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.58.mixer.k_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.58.mixer.v_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.58.mixer.o_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.59.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.59.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.59.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.59.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.59.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.59.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.60.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.60.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.61.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.61.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.61.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.61.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.61.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.61.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.62.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.62.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.63.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.63.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.63.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.63.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.63.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.63.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.64.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.64.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.65.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.65.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.65.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.65.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.65.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.65.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.66.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.66.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.67.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.67.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.67.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.67.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.67.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.67.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.68.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.68.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.69.mixer.q_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.69.mixer.k_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.69.mixer.v_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.69.mixer.o_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.70.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.70.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.70.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.70.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.70.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.70.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.71.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.71.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.72.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.72.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.72.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.72.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.72.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.72.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.73.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.73.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.74.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.74.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.74.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.74.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.74.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.74.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.75.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.75.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.76.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.76.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.76.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.76.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.76.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.76.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.77.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.77.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.78.mixer.q_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.78.mixer.k_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.78.mixer.v_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.78.mixer.o_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.79.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.79.mixer.switch_mlp.fc2": { "bits": 4, "group_size": 64 }, "backbone.layers.79.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.79.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.79.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.79.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.80.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.80.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.81.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.81.mixer.switch_mlp.fc2": { "bits": 4, "group_size": 64 }, "backbone.layers.81.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.81.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.81.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.81.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.82.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.82.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.83.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.83.mixer.switch_mlp.fc2": { "bits": 4, "group_size": 64 }, "backbone.layers.83.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.83.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.83.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.83.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.84.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.84.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.85.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.85.mixer.switch_mlp.fc2": { "bits": 4, "group_size": 64 }, "backbone.layers.85.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.85.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.85.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.85.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.86.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.86.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.87.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.87.mixer.switch_mlp.fc2": { "bits": 4, "group_size": 64 }, "backbone.layers.87.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.87.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.87.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.87.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "lm_head": { "bits": 4, "group_size": 64 } }, "quantization_config": { "group_size": 64, "bits": 4, "mode": "affine", "backbone.embeddings": { "bits": 4, "group_size": 64 }, "backbone.layers.0.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.0.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.1.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.1.mixer.switch_mlp.fc2": { "bits": 4, "group_size": 64 }, "backbone.layers.1.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.1.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.1.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.1.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.2.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.2.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.3.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.3.mixer.switch_mlp.fc2": { "bits": 4, "group_size": 64 }, "backbone.layers.3.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.3.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.3.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.3.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.4.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.4.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.5.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.5.mixer.switch_mlp.fc2": { "bits": 4, "group_size": 64 }, "backbone.layers.5.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.5.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.5.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.5.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.6.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.6.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.7.mixer.q_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.7.mixer.k_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.7.mixer.v_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.7.mixer.o_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.8.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.8.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.8.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.8.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.8.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.8.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.9.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.9.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.10.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.10.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.10.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.10.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.10.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.10.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.11.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.11.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.12.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.12.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.12.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.12.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.12.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.12.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.13.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.13.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.14.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.14.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.14.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.14.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.14.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.14.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.15.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.15.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.16.mixer.q_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.16.mixer.k_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.16.mixer.v_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.16.mixer.o_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.17.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.17.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.17.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.17.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.17.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.17.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.18.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.18.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.19.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.19.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.19.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.19.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.19.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.19.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.20.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.20.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.21.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.21.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.21.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.21.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.21.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.21.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.22.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.22.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.23.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.23.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.23.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.23.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.23.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.23.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.24.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.24.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.25.mixer.q_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.25.mixer.k_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.25.mixer.v_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.25.mixer.o_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.26.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.26.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.26.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.26.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.26.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.26.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.27.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.27.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.28.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.28.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.28.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.28.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.28.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.28.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.29.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.29.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.30.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.30.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.30.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.30.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.30.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.30.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.31.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.31.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.32.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.32.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.32.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.32.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.32.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.32.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.33.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.33.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.34.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.34.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.34.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.34.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.34.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.34.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.35.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.35.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.36.mixer.q_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.36.mixer.k_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.36.mixer.v_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.36.mixer.o_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.37.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.37.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.37.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.37.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.37.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.37.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.38.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.38.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.39.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.39.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.39.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.39.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.39.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.39.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.40.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.40.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.41.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.41.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.41.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.41.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.41.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.41.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.42.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.42.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.43.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.43.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.43.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.43.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.43.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.43.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.44.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.44.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.45.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.45.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.45.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.45.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.45.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.45.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.46.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.46.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.47.mixer.q_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.47.mixer.k_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.47.mixer.v_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.47.mixer.o_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.48.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.48.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.48.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.48.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.48.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.48.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.49.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.49.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.50.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.50.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.50.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.50.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.50.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.50.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.51.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.51.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.52.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.52.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.52.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.52.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.52.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.52.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.53.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.53.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.54.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.54.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.54.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.54.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.54.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.54.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.55.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.55.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.56.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.56.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.56.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.56.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.56.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.56.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.57.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.57.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.58.mixer.q_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.58.mixer.k_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.58.mixer.v_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.58.mixer.o_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.59.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.59.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.59.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.59.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.59.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.59.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.60.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.60.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.61.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.61.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.61.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.61.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.61.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.61.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.62.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.62.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.63.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.63.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.63.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.63.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.63.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.63.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.64.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.64.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.65.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.65.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.65.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.65.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.65.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.65.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.66.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.66.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.67.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.67.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.67.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.67.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.67.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.67.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.68.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.68.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.69.mixer.q_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.69.mixer.k_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.69.mixer.v_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.69.mixer.o_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.70.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.70.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.70.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.70.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.70.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.70.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.71.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.71.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.72.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.72.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.72.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.72.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.72.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.72.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.73.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.73.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.74.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.74.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.74.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.74.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.74.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.74.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.75.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.75.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.76.mixer.switch_mlp.fc1": { "bits": 2, "group_size": 64 }, "backbone.layers.76.mixer.switch_mlp.fc2": { "bits": 2, "group_size": 64 }, "backbone.layers.76.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.76.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.76.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.76.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.77.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.77.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.78.mixer.q_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.78.mixer.k_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.78.mixer.v_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.78.mixer.o_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.79.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.79.mixer.switch_mlp.fc2": { "bits": 4, "group_size": 64 }, "backbone.layers.79.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.79.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.79.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.79.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.80.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.80.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.81.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.81.mixer.switch_mlp.fc2": { "bits": 4, "group_size": 64 }, "backbone.layers.81.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.81.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.81.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.81.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.82.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.82.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.83.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.83.mixer.switch_mlp.fc2": { "bits": 4, "group_size": 64 }, "backbone.layers.83.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.83.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.83.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.83.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.84.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.84.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.85.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.85.mixer.switch_mlp.fc2": { "bits": 4, "group_size": 64 }, "backbone.layers.85.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.85.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.85.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.85.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.86.mixer.in_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.86.mixer.out_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.87.mixer.switch_mlp.fc1": { "bits": 4, "group_size": 64 }, "backbone.layers.87.mixer.switch_mlp.fc2": { "bits": 4, "group_size": 64 }, "backbone.layers.87.mixer.shared_experts.up_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.87.mixer.shared_experts.down_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.87.mixer.fc1_latent_proj": { "bits": 4, "group_size": 64 }, "backbone.layers.87.mixer.fc2_latent_proj": { "bits": 4, "group_size": 64 }, "lm_head": { "bits": 4, "group_size": 64 } }, "rescale_prenorm_residual": true, "residual_in_fp32": false, "rope_theta": 10000, "routed_scaling_factor": 5.0, "sliding_window": null, "ssm_state_size": 128, "tie_word_embeddings": false, "time_step_floor": 0.0001, "time_step_max": 0.1, "time_step_min": 0.001, "topk_group": 1, "transformers_version": "4.57.6", "use_bias": false, "use_cache": true, "use_conv_bias": true, "use_mamba_kernels": true, "vocab_size": 131072, "mtp_file": "mtp.safetensors", "mtp_tensor_count": 3108, "mtp_policy": "optiq-int4-prequantized-gs64", "mtplx_mtp_quantization": { "bits": 4, "group_size": 64, "mode": "affine", "policy": "cyankiwi", "prequantized": true }, "mtp_num_hidden_layers": 2 }