{
    "architectures": [
        "DeepseekV2ForCausalLM"
    ],
    "attention_bias": false,
    "attention_dropout": 0.0,
    "auto_map": {
        "AutoConfig": "configuration_deepseek.DeepseekV2Config",
        "AutoModel": "modeling_deepseek.DeepseekV2Model",
        "AutoModelForCausalLM": "modeling_deepseek.DeepseekV2ForCausalLM"
    },
    "aux_loss_alpha": 0.001,
    "bos_token_id": 100000,
    "eos_token_id": 100001,
    "first_k_dense_replace": 1,
    "hidden_act": "silu",
    "hidden_size": 2048,
    "initializer_range": 0.02,
    "intermediate_size": 10944,
    "kv_lora_rank": 512,
    "max_position_embeddings": 163840,
    "model_type": "deepseek_v2",
    "moe_intermediate_size": 1408,
    "moe_layer_freq": 1,
    "n_group": 1,
    "n_routed_experts": 64,
    "n_shared_experts": 2,
    "norm_topk_prob": false,
    "num_attention_heads": 16,
    "num_experts_per_tok": 6,
    "num_hidden_layers": 27,
    "num_key_value_heads": 16,
    "pretraining_tp": 1,
    "q_lora_rank": null,
    "qk_nope_head_dim": 128,
    "qk_rope_head_dim": 64,
    "quantization": {
        "group_size": 64,
        "bits": 4,
        "model.embed_tokens": {
            "group_size": 32,
            "bits": 4
        },
        "model.layers.0.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.0.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.0.self_attn.kv_a_layernorm": false,
        "model.layers.0.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.0.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.0.self_attn.rope": false,
        "model.layers.0.mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.0.mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.0.mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.0.input_layernorm": false,
        "model.layers.0.post_attention_layernorm": false,
        "model.layers.1.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.1.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.1.self_attn.kv_a_layernorm": false,
        "model.layers.1.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.1.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.1.self_attn.rope": false,
        "model.layers.1.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.1.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.1.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.1.mlp.switch_mlp.activation": false,
        "model.layers.1.mlp.gate": false,
        "model.layers.1.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.1.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.1.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.1.input_layernorm": false,
        "model.layers.1.post_attention_layernorm": false,
        "model.layers.2.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.2.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.2.self_attn.kv_a_layernorm": false,
        "model.layers.2.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.2.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.2.self_attn.rope": false,
        "model.layers.2.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.2.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.2.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.2.mlp.switch_mlp.activation": false,
        "model.layers.2.mlp.gate": false,
        "model.layers.2.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.2.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.2.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.2.input_layernorm": false,
        "model.layers.2.post_attention_layernorm": false,
        "model.layers.3.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.3.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.3.self_attn.kv_a_layernorm": false,
        "model.layers.3.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.3.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.3.self_attn.rope": false,
        "model.layers.3.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.3.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.3.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.3.mlp.switch_mlp.activation": false,
        "model.layers.3.mlp.gate": false,
        "model.layers.3.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.3.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.3.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.3.input_layernorm": false,
        "model.layers.3.post_attention_layernorm": false,
        "model.layers.4.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.4.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.4.self_attn.kv_a_layernorm": false,
        "model.layers.4.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.4.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.4.self_attn.rope": false,
        "model.layers.4.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.4.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.4.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.4.mlp.switch_mlp.activation": false,
        "model.layers.4.mlp.gate": false,
        "model.layers.4.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.4.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.4.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.4.input_layernorm": false,
        "model.layers.4.post_attention_layernorm": false,
        "model.layers.5.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.5.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.5.self_attn.kv_a_layernorm": false,
        "model.layers.5.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.5.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.5.self_attn.rope": false,
        "model.layers.5.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.5.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.5.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.5.mlp.switch_mlp.activation": false,
        "model.layers.5.mlp.gate": false,
        "model.layers.5.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.5.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.5.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.5.input_layernorm": false,
        "model.layers.5.post_attention_layernorm": false,
        "model.layers.6.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.6.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.6.self_attn.kv_a_layernorm": false,
        "model.layers.6.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.6.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.6.self_attn.rope": false,
        "model.layers.6.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.6.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.6.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.6.mlp.switch_mlp.activation": false,
        "model.layers.6.mlp.gate": false,
        "model.layers.6.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.6.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.6.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.6.input_layernorm": false,
        "model.layers.6.post_attention_layernorm": false,
        "model.layers.7.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.7.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.7.self_attn.kv_a_layernorm": false,
        "model.layers.7.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.7.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.7.self_attn.rope": false,
        "model.layers.7.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.7.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.7.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.7.mlp.switch_mlp.activation": false,
        "model.layers.7.mlp.gate": false,
        "model.layers.7.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.7.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.7.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.7.input_layernorm": false,
        "model.layers.7.post_attention_layernorm": false,
        "model.layers.8.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.8.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.8.self_attn.kv_a_layernorm": false,
        "model.layers.8.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.8.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.8.self_attn.rope": false,
        "model.layers.8.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.8.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.8.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.8.mlp.switch_mlp.activation": false,
        "model.layers.8.mlp.gate": false,
        "model.layers.8.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.8.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.8.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.8.input_layernorm": false,
        "model.layers.8.post_attention_layernorm": false,
        "model.layers.9.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.9.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.9.self_attn.kv_a_layernorm": false,
        "model.layers.9.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.9.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.9.self_attn.rope": false,
        "model.layers.9.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.9.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.9.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.9.mlp.switch_mlp.activation": false,
        "model.layers.9.mlp.gate": false,
        "model.layers.9.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.9.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.9.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.9.input_layernorm": false,
        "model.layers.9.post_attention_layernorm": false,
        "model.layers.10.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.10.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.10.self_attn.kv_a_layernorm": false,
        "model.layers.10.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.10.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.10.self_attn.rope": false,
        "model.layers.10.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.10.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.10.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.10.mlp.switch_mlp.activation": false,
        "model.layers.10.mlp.gate": false,
        "model.layers.10.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.10.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.10.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.10.input_layernorm": false,
        "model.layers.10.post_attention_layernorm": false,
        "model.layers.11.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.11.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.11.self_attn.kv_a_layernorm": false,
        "model.layers.11.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.11.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.11.self_attn.rope": false,
        "model.layers.11.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.11.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.11.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.11.mlp.switch_mlp.activation": false,
        "model.layers.11.mlp.gate": false,
        "model.layers.11.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.11.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.11.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.11.input_layernorm": false,
        "model.layers.11.post_attention_layernorm": false,
        "model.layers.12.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.12.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.12.self_attn.kv_a_layernorm": false,
        "model.layers.12.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.12.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.12.self_attn.rope": false,
        "model.layers.12.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.12.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.12.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.12.mlp.switch_mlp.activation": false,
        "model.layers.12.mlp.gate": false,
        "model.layers.12.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.12.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.12.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.12.input_layernorm": false,
        "model.layers.12.post_attention_layernorm": false,
        "model.layers.13.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.13.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.13.self_attn.kv_a_layernorm": false,
        "model.layers.13.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.13.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.13.self_attn.rope": false,
        "model.layers.13.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.13.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.13.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.13.mlp.switch_mlp.activation": false,
        "model.layers.13.mlp.gate": false,
        "model.layers.13.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.13.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.13.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.13.input_layernorm": false,
        "model.layers.13.post_attention_layernorm": false,
        "model.layers.14.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.14.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.14.self_attn.kv_a_layernorm": false,
        "model.layers.14.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.14.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.14.self_attn.rope": false,
        "model.layers.14.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.14.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.14.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.14.mlp.switch_mlp.activation": false,
        "model.layers.14.mlp.gate": false,
        "model.layers.14.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.14.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.14.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.14.input_layernorm": false,
        "model.layers.14.post_attention_layernorm": false,
        "model.layers.15.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.15.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.15.self_attn.kv_a_layernorm": false,
        "model.layers.15.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.15.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.15.self_attn.rope": false,
        "model.layers.15.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.15.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.15.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.15.mlp.switch_mlp.activation": false,
        "model.layers.15.mlp.gate": false,
        "model.layers.15.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.15.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.15.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.15.input_layernorm": false,
        "model.layers.15.post_attention_layernorm": false,
        "model.layers.16.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.16.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.16.self_attn.kv_a_layernorm": false,
        "model.layers.16.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.16.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.16.self_attn.rope": false,
        "model.layers.16.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.16.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.16.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.16.mlp.switch_mlp.activation": false,
        "model.layers.16.mlp.gate": false,
        "model.layers.16.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.16.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.16.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.16.input_layernorm": false,
        "model.layers.16.post_attention_layernorm": false,
        "model.layers.17.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.17.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.17.self_attn.kv_a_layernorm": false,
        "model.layers.17.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.17.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.17.self_attn.rope": false,
        "model.layers.17.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.17.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.17.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.17.mlp.switch_mlp.activation": false,
        "model.layers.17.mlp.gate": false,
        "model.layers.17.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.17.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.17.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.17.input_layernorm": false,
        "model.layers.17.post_attention_layernorm": false,
        "model.layers.18.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.18.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.18.self_attn.kv_a_layernorm": false,
        "model.layers.18.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.18.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.18.self_attn.rope": false,
        "model.layers.18.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.18.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.18.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.18.mlp.switch_mlp.activation": false,
        "model.layers.18.mlp.gate": false,
        "model.layers.18.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.18.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.18.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.18.input_layernorm": false,
        "model.layers.18.post_attention_layernorm": false,
        "model.layers.19.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.19.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.19.self_attn.kv_a_layernorm": false,
        "model.layers.19.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.19.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.19.self_attn.rope": false,
        "model.layers.19.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.19.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.19.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.19.mlp.switch_mlp.activation": false,
        "model.layers.19.mlp.gate": false,
        "model.layers.19.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.19.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.19.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.19.input_layernorm": false,
        "model.layers.19.post_attention_layernorm": false,
        "model.layers.20.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.20.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.20.self_attn.kv_a_layernorm": false,
        "model.layers.20.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.20.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.20.self_attn.rope": false,
        "model.layers.20.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.20.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.20.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.20.mlp.switch_mlp.activation": false,
        "model.layers.20.mlp.gate": false,
        "model.layers.20.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.20.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.20.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.20.input_layernorm": false,
        "model.layers.20.post_attention_layernorm": false,
        "model.layers.21.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.21.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.21.self_attn.kv_a_layernorm": false,
        "model.layers.21.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.21.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.21.self_attn.rope": false,
        "model.layers.21.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.21.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.21.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.21.mlp.switch_mlp.activation": false,
        "model.layers.21.mlp.gate": false,
        "model.layers.21.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.21.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.21.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.21.input_layernorm": false,
        "model.layers.21.post_attention_layernorm": false,
        "model.layers.22.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.22.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.22.self_attn.kv_a_layernorm": false,
        "model.layers.22.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.22.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.22.self_attn.rope": false,
        "model.layers.22.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.22.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.22.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.22.mlp.switch_mlp.activation": false,
        "model.layers.22.mlp.gate": false,
        "model.layers.22.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.22.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.22.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.22.input_layernorm": false,
        "model.layers.22.post_attention_layernorm": false,
        "model.layers.23.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.23.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.23.self_attn.kv_a_layernorm": false,
        "model.layers.23.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.23.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.23.self_attn.rope": false,
        "model.layers.23.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.23.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.23.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.23.mlp.switch_mlp.activation": false,
        "model.layers.23.mlp.gate": false,
        "model.layers.23.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.23.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.23.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.23.input_layernorm": false,
        "model.layers.23.post_attention_layernorm": false,
        "model.layers.24.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.24.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.24.self_attn.kv_a_layernorm": false,
        "model.layers.24.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.24.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.24.self_attn.rope": false,
        "model.layers.24.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.24.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.24.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.24.mlp.switch_mlp.activation": false,
        "model.layers.24.mlp.gate": false,
        "model.layers.24.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.24.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.24.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.24.input_layernorm": false,
        "model.layers.24.post_attention_layernorm": false,
        "model.layers.25.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.25.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.25.self_attn.kv_a_layernorm": false,
        "model.layers.25.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.25.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.25.self_attn.rope": false,
        "model.layers.25.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.25.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.25.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.25.mlp.switch_mlp.activation": false,
        "model.layers.25.mlp.gate": false,
        "model.layers.25.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.25.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.25.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.25.input_layernorm": false,
        "model.layers.25.post_attention_layernorm": false,
        "model.layers.26.self_attn.q_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.26.self_attn.kv_a_proj_with_mqa": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.26.self_attn.kv_a_layernorm": false,
        "model.layers.26.self_attn.kv_b_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.26.self_attn.o_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.26.self_attn.rope": false,
        "model.layers.26.mlp.switch_mlp.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.26.mlp.switch_mlp.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.26.mlp.switch_mlp.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.26.mlp.switch_mlp.activation": false,
        "model.layers.26.mlp.gate": false,
        "model.layers.26.mlp.shared_experts.gate_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.26.mlp.shared_experts.up_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.26.mlp.shared_experts.down_proj": {
            "group_size": 64,
            "bits": 4
        },
        "model.layers.26.input_layernorm": false,
        "model.layers.26.post_attention_layernorm": false,
        "model.norm": false,
        "lm_head": {
            "group_size": 32,
            "bits": 4
        }
    },
    "rms_norm_eps": 1e-06,
    "rope_scaling": {
        "beta_fast": 32,
        "beta_slow": 1,
        "factor": 40,
        "mscale": 0.707,
        "mscale_all_dim": 0.707,
        "original_max_position_embeddings": 4096,
        "type": "yarn"
    },
    "rope_theta": 10000,
    "routed_scaling_factor": 1.0,
    "scoring_func": "softmax",
    "seq_aux": true,
    "tie_word_embeddings": false,
    "topk_group": 1,
    "topk_method": "greedy",
    "torch_dtype": "bfloat16",
    "transformers_version": "4.39.3",
    "use_cache": true,
    "v_head_dim": 128,
    "vocab_size": 102400
}