diff --git "a/config.json" "b/config.json" new file mode 100644--- /dev/null +++ "b/config.json" @@ -0,0 +1,3456 @@ +{ + "architectures": [ + "K2HorizonForCausalLM" + ], + "attention_bias": false, + "attention_dropout": 0.0, + "attention_gate_func": "softplus", + "auto_map": { + "AutoConfig": "configuration_k2_horizon.K2HorizonConfig", + "AutoModel": "modeling_k2_horizon.K2HorizonModel", + "AutoModelForCausalLM": "modeling_k2_horizon.K2HorizonForCausalLM" + }, + "bos_token_id": 0, + "decoder_sparse_step": 1, + "dtype": "bfloat16", + "eos_token_id": 1, + "head_dim": 128, + "hidden_act": "silu", + "hidden_size": 2560, + "initializer_range": 0.02, + "intermediate_size": 6144, + "layernorm_num_groups": 2, + "max_position_embeddings": 524288, + "mlp_only_layers": [ + 0, + 1, + 2 + ], + "model_type": "k2_horizon", + "moe_gate_bias": true, + "moe_intermediate_size": 768, + "mova_num_experts": 64, + "mova_num_experts_per_tok": 4, + "norm_topk_prob": true, + "num_attention_heads": 32, + "num_experts": 100, + "num_experts_per_tok": 8, + "num_hidden_layers": 48, + "num_key_value_heads": 8, + "num_shared_experts": 1, + "output_router_logits": false, + "pad_token_id": null, + "quantization_config": { + "config_groups": { + "group_0": { + "format": "nvfp4-pack-quantized", + "input_activations": { + "actorder": null, + "block_structure": null, + "dynamic": "local", + "group_size": 16, + "num_bits": 4, + "observer": "static_minmax", + "observer_kwargs": {}, + "scale_dtype": "torch.float8_e4m3fn", + "strategy": "tensor_group", + "symmetric": true, + "type": "float", + "zp_dtype": null + }, + "output_activations": null, + "targets": [ + "re:.*\\.layers\\.(3|6|9|12|15|18|21|24|27|30|33|36|39|42|45)\\.mlp\\.experts\\.\\d+\\.(gate_proj|up_proj|down_proj)$" + ], + "weights": { + "actorder": null, + "block_structure": null, + "dynamic": false, + "group_size": 16, + "num_bits": 4, + "observer": "memoryless_minmax", + "observer_kwargs": {}, + "scale_dtype": "torch.float8_e4m3fn", + "strategy": "tensor_group", + "symmetric": true, + "type": "float", + "zp_dtype": null + } + }, + "group_1": { + "format": "float-quantized", + "input_activations": { + "actorder": null, + "block_structure": null, + "dynamic": true, + "group_size": 128, + "num_bits": 8, + "observer": null, + "observer_kwargs": {}, + "scale_dtype": null, + "strategy": "group", + "symmetric": true, + "type": "float", + "zp_dtype": null + }, + "output_activations": null, + "targets": [ + "re:.*\\.layers\\.(4|5|7|8|10|11|13|14|16|17|19|20|22|23|25|26|28|29|31|32|34|35|37|38|40|41|43|44|46|47)\\.mlp\\.experts\\.\\d+\\.(gate_proj|up_proj|down_proj)$" + ], + "weights": { + "actorder": null, + "block_structure": [ + 128, + 128 + ], + "dynamic": false, + "group_size": null, + "num_bits": 8, + "observer": "memoryless_minmax", + "observer_kwargs": {}, + "scale_dtype": null, + "strategy": "block", + "symmetric": true, + "type": "float", + "zp_dtype": null + } + } + }, + "format": "mixed-precision", + "global_compression_ratio": null, + "ignore": [ + "model.layers.0.self_attn.q_proj", + "model.layers.0.self_attn.k_proj", + "model.layers.0.self_attn.v_proj", + "model.layers.0.self_attn.o_proj", + "model.layers.0.self_attn.gate_proj", + "model.layers.0.mlp.gate_proj", + "model.layers.0.mlp.up_proj", + "model.layers.0.mlp.down_proj", + "model.layers.1.self_attn.q_proj", + "model.layers.1.self_attn.k_proj", + "model.layers.1.self_attn.v_proj", + "model.layers.1.self_attn.o_proj", + "model.layers.1.self_attn.gate_proj", + "model.layers.1.mlp.gate_proj", + "model.layers.1.mlp.up_proj", + "model.layers.1.mlp.down_proj", + "model.layers.2.self_attn.q_proj", + "model.layers.2.self_attn.k_proj", + "model.layers.2.self_attn.v_proj", + "model.layers.2.self_attn.o_proj", + "model.layers.2.self_attn.gate_proj", + "model.layers.2.mlp.gate_proj", + "model.layers.2.mlp.up_proj", + "model.layers.2.mlp.down_proj", + "model.layers.3.self_attn.q_proj", + "model.layers.3.self_attn.k_proj", + "model.layers.3.self_attn.o_proj", + "model.layers.3.self_attn.v_router", + "model.layers.3.self_attn.v_experts.0", + "model.layers.3.self_attn.v_experts.1", + "model.layers.3.self_attn.v_experts.2", + "model.layers.3.self_attn.v_experts.3", + "model.layers.3.self_attn.v_experts.4", + "model.layers.3.self_attn.v_experts.5", + "model.layers.3.self_attn.v_experts.6", + "model.layers.3.self_attn.v_experts.7", + "model.layers.3.self_attn.v_experts.8", + "model.layers.3.self_attn.v_experts.9", + "model.layers.3.self_attn.v_experts.10", + "model.layers.3.self_attn.v_experts.11", + "model.layers.3.self_attn.v_experts.12", + "model.layers.3.self_attn.v_experts.13", + "model.layers.3.self_attn.v_experts.14", + "model.layers.3.self_attn.v_experts.15", + "model.layers.3.self_attn.v_experts.16", + "model.layers.3.self_attn.v_experts.17", + "model.layers.3.self_attn.v_experts.18", + "model.layers.3.self_attn.v_experts.19", + "model.layers.3.self_attn.v_experts.20", + "model.layers.3.self_attn.v_experts.21", + "model.layers.3.self_attn.v_experts.22", + "model.layers.3.self_attn.v_experts.23", + "model.layers.3.self_attn.v_experts.24", + "model.layers.3.self_attn.v_experts.25", + "model.layers.3.self_attn.v_experts.26", + "model.layers.3.self_attn.v_experts.27", + "model.layers.3.self_attn.v_experts.28", + "model.layers.3.self_attn.v_experts.29", + "model.layers.3.self_attn.v_experts.30", + "model.layers.3.self_attn.v_experts.31", + "model.layers.3.self_attn.v_experts.32", + "model.layers.3.self_attn.v_experts.33", + "model.layers.3.self_attn.v_experts.34", + "model.layers.3.self_attn.v_experts.35", + "model.layers.3.self_attn.v_experts.36", + "model.layers.3.self_attn.v_experts.37", + "model.layers.3.self_attn.v_experts.38", + "model.layers.3.self_attn.v_experts.39", + "model.layers.3.self_attn.v_experts.40", + "model.layers.3.self_attn.v_experts.41", + "model.layers.3.self_attn.v_experts.42", + "model.layers.3.self_attn.v_experts.43", + "model.layers.3.self_attn.v_experts.44", + "model.layers.3.self_attn.v_experts.45", + "model.layers.3.self_attn.v_experts.46", + "model.layers.3.self_attn.v_experts.47", + "model.layers.3.self_attn.v_experts.48", + "model.layers.3.self_attn.v_experts.49", + "model.layers.3.self_attn.v_experts.50", + "model.layers.3.self_attn.v_experts.51", + "model.layers.3.self_attn.v_experts.52", + "model.layers.3.self_attn.v_experts.53", + "model.layers.3.self_attn.v_experts.54", + "model.layers.3.self_attn.v_experts.55", + "model.layers.3.self_attn.v_experts.56", + "model.layers.3.self_attn.v_experts.57", + "model.layers.3.self_attn.v_experts.58", + "model.layers.3.self_attn.v_experts.59", + "model.layers.3.self_attn.v_experts.60", + "model.layers.3.self_attn.v_experts.61", + "model.layers.3.self_attn.v_experts.62", + "model.layers.3.self_attn.v_experts.63", + "model.layers.3.self_attn.gate_proj", + "model.layers.3.mlp.gate", + "model.layers.3.mlp.shared_experts.gate_proj", + "model.layers.3.mlp.shared_experts.up_proj", + "model.layers.3.mlp.shared_experts.down_proj", + "model.layers.4.self_attn.q_proj", + "model.layers.4.self_attn.k_proj", + "model.layers.4.self_attn.o_proj", + "model.layers.4.self_attn.v_router", + "model.layers.4.self_attn.v_experts.0", + "model.layers.4.self_attn.v_experts.1", + "model.layers.4.self_attn.v_experts.2", + "model.layers.4.self_attn.v_experts.3", + "model.layers.4.self_attn.v_experts.4", + "model.layers.4.self_attn.v_experts.5", + "model.layers.4.self_attn.v_experts.6", + "model.layers.4.self_attn.v_experts.7", + "model.layers.4.self_attn.v_experts.8", + "model.layers.4.self_attn.v_experts.9", + "model.layers.4.self_attn.v_experts.10", + "model.layers.4.self_attn.v_experts.11", + "model.layers.4.self_attn.v_experts.12", + "model.layers.4.self_attn.v_experts.13", + "model.layers.4.self_attn.v_experts.14", + "model.layers.4.self_attn.v_experts.15", + "model.layers.4.self_attn.v_experts.16", + "model.layers.4.self_attn.v_experts.17", + "model.layers.4.self_attn.v_experts.18", + "model.layers.4.self_attn.v_experts.19", + "model.layers.4.self_attn.v_experts.20", + "model.layers.4.self_attn.v_experts.21", + "model.layers.4.self_attn.v_experts.22", + "model.layers.4.self_attn.v_experts.23", + "model.layers.4.self_attn.v_experts.24", + "model.layers.4.self_attn.v_experts.25", + "model.layers.4.self_attn.v_experts.26", + "model.layers.4.self_attn.v_experts.27", + "model.layers.4.self_attn.v_experts.28", + "model.layers.4.self_attn.v_experts.29", + "model.layers.4.self_attn.v_experts.30", + "model.layers.4.self_attn.v_experts.31", + "model.layers.4.self_attn.v_experts.32", + "model.layers.4.self_attn.v_experts.33", + "model.layers.4.self_attn.v_experts.34", + "model.layers.4.self_attn.v_experts.35", + "model.layers.4.self_attn.v_experts.36", + "model.layers.4.self_attn.v_experts.37", + "model.layers.4.self_attn.v_experts.38", + "model.layers.4.self_attn.v_experts.39", + "model.layers.4.self_attn.v_experts.40", + "model.layers.4.self_attn.v_experts.41", + "model.layers.4.self_attn.v_experts.42", + "model.layers.4.self_attn.v_experts.43", + "model.layers.4.self_attn.v_experts.44", + "model.layers.4.self_attn.v_experts.45", + "model.layers.4.self_attn.v_experts.46", + "model.layers.4.self_attn.v_experts.47", + "model.layers.4.self_attn.v_experts.48", + "model.layers.4.self_attn.v_experts.49", + "model.layers.4.self_attn.v_experts.50", + "model.layers.4.self_attn.v_experts.51", + "model.layers.4.self_attn.v_experts.52", + "model.layers.4.self_attn.v_experts.53", + "model.layers.4.self_attn.v_experts.54", + "model.layers.4.self_attn.v_experts.55", + "model.layers.4.self_attn.v_experts.56", + "model.layers.4.self_attn.v_experts.57", + "model.layers.4.self_attn.v_experts.58", + "model.layers.4.self_attn.v_experts.59", + "model.layers.4.self_attn.v_experts.60", + "model.layers.4.self_attn.v_experts.61", + "model.layers.4.self_attn.v_experts.62", + "model.layers.4.self_attn.v_experts.63", + "model.layers.4.self_attn.gate_proj", + "model.layers.4.mlp.gate", + "model.layers.4.mlp.shared_experts.gate_proj", + "model.layers.4.mlp.shared_experts.up_proj", + "model.layers.4.mlp.shared_experts.down_proj", + "model.layers.5.self_attn.q_proj", + "model.layers.5.self_attn.k_proj", + "model.layers.5.self_attn.o_proj", + "model.layers.5.self_attn.v_router", + "model.layers.5.self_attn.v_experts.0", + "model.layers.5.self_attn.v_experts.1", + "model.layers.5.self_attn.v_experts.2", + "model.layers.5.self_attn.v_experts.3", + "model.layers.5.self_attn.v_experts.4", + "model.layers.5.self_attn.v_experts.5", + "model.layers.5.self_attn.v_experts.6", + "model.layers.5.self_attn.v_experts.7", + "model.layers.5.self_attn.v_experts.8", + "model.layers.5.self_attn.v_experts.9", + "model.layers.5.self_attn.v_experts.10", + "model.layers.5.self_attn.v_experts.11", + "model.layers.5.self_attn.v_experts.12", + "model.layers.5.self_attn.v_experts.13", + "model.layers.5.self_attn.v_experts.14", + "model.layers.5.self_attn.v_experts.15", + "model.layers.5.self_attn.v_experts.16", + "model.layers.5.self_attn.v_experts.17", + "model.layers.5.self_attn.v_experts.18", + "model.layers.5.self_attn.v_experts.19", + "model.layers.5.self_attn.v_experts.20", + "model.layers.5.self_attn.v_experts.21", + "model.layers.5.self_attn.v_experts.22", + "model.layers.5.self_attn.v_experts.23", + "model.layers.5.self_attn.v_experts.24", + "model.layers.5.self_attn.v_experts.25", + "model.layers.5.self_attn.v_experts.26", + "model.layers.5.self_attn.v_experts.27", + "model.layers.5.self_attn.v_experts.28", + "model.layers.5.self_attn.v_experts.29", + "model.layers.5.self_attn.v_experts.30", + "model.layers.5.self_attn.v_experts.31", + "model.layers.5.self_attn.v_experts.32", + "model.layers.5.self_attn.v_experts.33", + "model.layers.5.self_attn.v_experts.34", + "model.layers.5.self_attn.v_experts.35", + "model.layers.5.self_attn.v_experts.36", + "model.layers.5.self_attn.v_experts.37", + "model.layers.5.self_attn.v_experts.38", + "model.layers.5.self_attn.v_experts.39", + "model.layers.5.self_attn.v_experts.40", + "model.layers.5.self_attn.v_experts.41", + "model.layers.5.self_attn.v_experts.42", + "model.layers.5.self_attn.v_experts.43", + "model.layers.5.self_attn.v_experts.44", + "model.layers.5.self_attn.v_experts.45", + "model.layers.5.self_attn.v_experts.46", + "model.layers.5.self_attn.v_experts.47", + "model.layers.5.self_attn.v_experts.48", + "model.layers.5.self_attn.v_experts.49", + "model.layers.5.self_attn.v_experts.50", + "model.layers.5.self_attn.v_experts.51", + "model.layers.5.self_attn.v_experts.52", + "model.layers.5.self_attn.v_experts.53", + "model.layers.5.self_attn.v_experts.54", + "model.layers.5.self_attn.v_experts.55", + "model.layers.5.self_attn.v_experts.56", + "model.layers.5.self_attn.v_experts.57", + "model.layers.5.self_attn.v_experts.58", + "model.layers.5.self_attn.v_experts.59", + "model.layers.5.self_attn.v_experts.60", + "model.layers.5.self_attn.v_experts.61", + "model.layers.5.self_attn.v_experts.62", + "model.layers.5.self_attn.v_experts.63", + "model.layers.5.self_attn.gate_proj", + "model.layers.5.mlp.gate", + "model.layers.5.mlp.shared_experts.gate_proj", + "model.layers.5.mlp.shared_experts.up_proj", + "model.layers.5.mlp.shared_experts.down_proj", + "model.layers.6.self_attn.q_proj", + "model.layers.6.self_attn.k_proj", + "model.layers.6.self_attn.o_proj", + "model.layers.6.self_attn.v_router", + "model.layers.6.self_attn.v_experts.0", + "model.layers.6.self_attn.v_experts.1", + "model.layers.6.self_attn.v_experts.2", + "model.layers.6.self_attn.v_experts.3", + "model.layers.6.self_attn.v_experts.4", + "model.layers.6.self_attn.v_experts.5", + "model.layers.6.self_attn.v_experts.6", + "model.layers.6.self_attn.v_experts.7", + "model.layers.6.self_attn.v_experts.8", + "model.layers.6.self_attn.v_experts.9", + "model.layers.6.self_attn.v_experts.10", + "model.layers.6.self_attn.v_experts.11", + "model.layers.6.self_attn.v_experts.12", + "model.layers.6.self_attn.v_experts.13", + "model.layers.6.self_attn.v_experts.14", + "model.layers.6.self_attn.v_experts.15", + "model.layers.6.self_attn.v_experts.16", + "model.layers.6.self_attn.v_experts.17", + "model.layers.6.self_attn.v_experts.18", + "model.layers.6.self_attn.v_experts.19", + "model.layers.6.self_attn.v_experts.20", + "model.layers.6.self_attn.v_experts.21", + "model.layers.6.self_attn.v_experts.22", + "model.layers.6.self_attn.v_experts.23", + "model.layers.6.self_attn.v_experts.24", + "model.layers.6.self_attn.v_experts.25", + "model.layers.6.self_attn.v_experts.26", + "model.layers.6.self_attn.v_experts.27", + "model.layers.6.self_attn.v_experts.28", + "model.layers.6.self_attn.v_experts.29", + "model.layers.6.self_attn.v_experts.30", + "model.layers.6.self_attn.v_experts.31", + "model.layers.6.self_attn.v_experts.32", + "model.layers.6.self_attn.v_experts.33", + "model.layers.6.self_attn.v_experts.34", + "model.layers.6.self_attn.v_experts.35", + "model.layers.6.self_attn.v_experts.36", + "model.layers.6.self_attn.v_experts.37", + "model.layers.6.self_attn.v_experts.38", + "model.layers.6.self_attn.v_experts.39", + "model.layers.6.self_attn.v_experts.40", + "model.layers.6.self_attn.v_experts.41", + "model.layers.6.self_attn.v_experts.42", + "model.layers.6.self_attn.v_experts.43", + "model.layers.6.self_attn.v_experts.44", + "model.layers.6.self_attn.v_experts.45", + "model.layers.6.self_attn.v_experts.46", + "model.layers.6.self_attn.v_experts.47", + "model.layers.6.self_attn.v_experts.48", + "model.layers.6.self_attn.v_experts.49", + "model.layers.6.self_attn.v_experts.50", + "model.layers.6.self_attn.v_experts.51", + "model.layers.6.self_attn.v_experts.52", + "model.layers.6.self_attn.v_experts.53", + "model.layers.6.self_attn.v_experts.54", + "model.layers.6.self_attn.v_experts.55", + "model.layers.6.self_attn.v_experts.56", + "model.layers.6.self_attn.v_experts.57", + "model.layers.6.self_attn.v_experts.58", + "model.layers.6.self_attn.v_experts.59", + "model.layers.6.self_attn.v_experts.60", + "model.layers.6.self_attn.v_experts.61", + "model.layers.6.self_attn.v_experts.62", + "model.layers.6.self_attn.v_experts.63", + "model.layers.6.self_attn.gate_proj", + "model.layers.6.mlp.gate", + "model.layers.6.mlp.shared_experts.gate_proj", + "model.layers.6.mlp.shared_experts.up_proj", + "model.layers.6.mlp.shared_experts.down_proj", + "model.layers.7.self_attn.q_proj", + "model.layers.7.self_attn.k_proj", + "model.layers.7.self_attn.o_proj", + "model.layers.7.self_attn.v_router", + "model.layers.7.self_attn.v_experts.0", + "model.layers.7.self_attn.v_experts.1", + "model.layers.7.self_attn.v_experts.2", + "model.layers.7.self_attn.v_experts.3", + "model.layers.7.self_attn.v_experts.4", + "model.layers.7.self_attn.v_experts.5", + "model.layers.7.self_attn.v_experts.6", + "model.layers.7.self_attn.v_experts.7", + "model.layers.7.self_attn.v_experts.8", + "model.layers.7.self_attn.v_experts.9", + "model.layers.7.self_attn.v_experts.10", + "model.layers.7.self_attn.v_experts.11", + "model.layers.7.self_attn.v_experts.12", + "model.layers.7.self_attn.v_experts.13", + "model.layers.7.self_attn.v_experts.14", + "model.layers.7.self_attn.v_experts.15", + "model.layers.7.self_attn.v_experts.16", + "model.layers.7.self_attn.v_experts.17", + "model.layers.7.self_attn.v_experts.18", + "model.layers.7.self_attn.v_experts.19", + "model.layers.7.self_attn.v_experts.20", + "model.layers.7.self_attn.v_experts.21", + "model.layers.7.self_attn.v_experts.22", + "model.layers.7.self_attn.v_experts.23", + "model.layers.7.self_attn.v_experts.24", + "model.layers.7.self_attn.v_experts.25", + "model.layers.7.self_attn.v_experts.26", + "model.layers.7.self_attn.v_experts.27", + "model.layers.7.self_attn.v_experts.28", + "model.layers.7.self_attn.v_experts.29", + "model.layers.7.self_attn.v_experts.30", + "model.layers.7.self_attn.v_experts.31", + "model.layers.7.self_attn.v_experts.32", + "model.layers.7.self_attn.v_experts.33", + "model.layers.7.self_attn.v_experts.34", + "model.layers.7.self_attn.v_experts.35", + "model.layers.7.self_attn.v_experts.36", + "model.layers.7.self_attn.v_experts.37", + "model.layers.7.self_attn.v_experts.38", + "model.layers.7.self_attn.v_experts.39", + "model.layers.7.self_attn.v_experts.40", + "model.layers.7.self_attn.v_experts.41", + "model.layers.7.self_attn.v_experts.42", + "model.layers.7.self_attn.v_experts.43", + "model.layers.7.self_attn.v_experts.44", + "model.layers.7.self_attn.v_experts.45", + "model.layers.7.self_attn.v_experts.46", + "model.layers.7.self_attn.v_experts.47", + "model.layers.7.self_attn.v_experts.48", + "model.layers.7.self_attn.v_experts.49", + "model.layers.7.self_attn.v_experts.50", + "model.layers.7.self_attn.v_experts.51", + "model.layers.7.self_attn.v_experts.52", + "model.layers.7.self_attn.v_experts.53", + "model.layers.7.self_attn.v_experts.54", + "model.layers.7.self_attn.v_experts.55", + "model.layers.7.self_attn.v_experts.56", + "model.layers.7.self_attn.v_experts.57", + "model.layers.7.self_attn.v_experts.58", + "model.layers.7.self_attn.v_experts.59", + "model.layers.7.self_attn.v_experts.60", + "model.layers.7.self_attn.v_experts.61", + "model.layers.7.self_attn.v_experts.62", + "model.layers.7.self_attn.v_experts.63", + "model.layers.7.self_attn.gate_proj", + "model.layers.7.mlp.gate", + "model.layers.7.mlp.shared_experts.gate_proj", + "model.layers.7.mlp.shared_experts.up_proj", + "model.layers.7.mlp.shared_experts.down_proj", + "model.layers.8.self_attn.q_proj", + "model.layers.8.self_attn.k_proj", + "model.layers.8.self_attn.o_proj", + "model.layers.8.self_attn.v_router", + "model.layers.8.self_attn.v_experts.0", + "model.layers.8.self_attn.v_experts.1", + "model.layers.8.self_attn.v_experts.2", + "model.layers.8.self_attn.v_experts.3", + "model.layers.8.self_attn.v_experts.4", + "model.layers.8.self_attn.v_experts.5", + "model.layers.8.self_attn.v_experts.6", + "model.layers.8.self_attn.v_experts.7", + "model.layers.8.self_attn.v_experts.8", + "model.layers.8.self_attn.v_experts.9", + "model.layers.8.self_attn.v_experts.10", + "model.layers.8.self_attn.v_experts.11", + "model.layers.8.self_attn.v_experts.12", + "model.layers.8.self_attn.v_experts.13", + "model.layers.8.self_attn.v_experts.14", + "model.layers.8.self_attn.v_experts.15", + "model.layers.8.self_attn.v_experts.16", + "model.layers.8.self_attn.v_experts.17", + "model.layers.8.self_attn.v_experts.18", + "model.layers.8.self_attn.v_experts.19", + "model.layers.8.self_attn.v_experts.20", + "model.layers.8.self_attn.v_experts.21", + "model.layers.8.self_attn.v_experts.22", + "model.layers.8.self_attn.v_experts.23", + "model.layers.8.self_attn.v_experts.24", + "model.layers.8.self_attn.v_experts.25", + "model.layers.8.self_attn.v_experts.26", + "model.layers.8.self_attn.v_experts.27", + "model.layers.8.self_attn.v_experts.28", + "model.layers.8.self_attn.v_experts.29", + "model.layers.8.self_attn.v_experts.30", + "model.layers.8.self_attn.v_experts.31", + "model.layers.8.self_attn.v_experts.32", + "model.layers.8.self_attn.v_experts.33", + "model.layers.8.self_attn.v_experts.34", + "model.layers.8.self_attn.v_experts.35", + "model.layers.8.self_attn.v_experts.36", + "model.layers.8.self_attn.v_experts.37", + "model.layers.8.self_attn.v_experts.38", + "model.layers.8.self_attn.v_experts.39", + "model.layers.8.self_attn.v_experts.40", + "model.layers.8.self_attn.v_experts.41", + "model.layers.8.self_attn.v_experts.42", + "model.layers.8.self_attn.v_experts.43", + "model.layers.8.self_attn.v_experts.44", + "model.layers.8.self_attn.v_experts.45", + "model.layers.8.self_attn.v_experts.46", + "model.layers.8.self_attn.v_experts.47", + "model.layers.8.self_attn.v_experts.48", + "model.layers.8.self_attn.v_experts.49", + "model.layers.8.self_attn.v_experts.50", + "model.layers.8.self_attn.v_experts.51", + "model.layers.8.self_attn.v_experts.52", + "model.layers.8.self_attn.v_experts.53", + "model.layers.8.self_attn.v_experts.54", + "model.layers.8.self_attn.v_experts.55", + "model.layers.8.self_attn.v_experts.56", + "model.layers.8.self_attn.v_experts.57", + "model.layers.8.self_attn.v_experts.58", + "model.layers.8.self_attn.v_experts.59", + "model.layers.8.self_attn.v_experts.60", + "model.layers.8.self_attn.v_experts.61", + "model.layers.8.self_attn.v_experts.62", + "model.layers.8.self_attn.v_experts.63", + "model.layers.8.self_attn.gate_proj", + "model.layers.8.mlp.gate", + "model.layers.8.mlp.shared_experts.gate_proj", + "model.layers.8.mlp.shared_experts.up_proj", + "model.layers.8.mlp.shared_experts.down_proj", + "model.layers.9.self_attn.q_proj", + "model.layers.9.self_attn.k_proj", + "model.layers.9.self_attn.o_proj", + "model.layers.9.self_attn.v_router", + "model.layers.9.self_attn.v_experts.0", + "model.layers.9.self_attn.v_experts.1", + "model.layers.9.self_attn.v_experts.2", + "model.layers.9.self_attn.v_experts.3", + "model.layers.9.self_attn.v_experts.4", + "model.layers.9.self_attn.v_experts.5", + "model.layers.9.self_attn.v_experts.6", + "model.layers.9.self_attn.v_experts.7", + "model.layers.9.self_attn.v_experts.8", + "model.layers.9.self_attn.v_experts.9", + "model.layers.9.self_attn.v_experts.10", + "model.layers.9.self_attn.v_experts.11", + "model.layers.9.self_attn.v_experts.12", + "model.layers.9.self_attn.v_experts.13", + "model.layers.9.self_attn.v_experts.14", + "model.layers.9.self_attn.v_experts.15", + "model.layers.9.self_attn.v_experts.16", + "model.layers.9.self_attn.v_experts.17", + "model.layers.9.self_attn.v_experts.18", + "model.layers.9.self_attn.v_experts.19", + "model.layers.9.self_attn.v_experts.20", + "model.layers.9.self_attn.v_experts.21", + "model.layers.9.self_attn.v_experts.22", + "model.layers.9.self_attn.v_experts.23", + "model.layers.9.self_attn.v_experts.24", + "model.layers.9.self_attn.v_experts.25", + "model.layers.9.self_attn.v_experts.26", + "model.layers.9.self_attn.v_experts.27", + "model.layers.9.self_attn.v_experts.28", + "model.layers.9.self_attn.v_experts.29", + "model.layers.9.self_attn.v_experts.30", + "model.layers.9.self_attn.v_experts.31", + "model.layers.9.self_attn.v_experts.32", + "model.layers.9.self_attn.v_experts.33", + "model.layers.9.self_attn.v_experts.34", + "model.layers.9.self_attn.v_experts.35", + "model.layers.9.self_attn.v_experts.36", + "model.layers.9.self_attn.v_experts.37", + "model.layers.9.self_attn.v_experts.38", + "model.layers.9.self_attn.v_experts.39", + "model.layers.9.self_attn.v_experts.40", + "model.layers.9.self_attn.v_experts.41", + "model.layers.9.self_attn.v_experts.42", + "model.layers.9.self_attn.v_experts.43", + "model.layers.9.self_attn.v_experts.44", + "model.layers.9.self_attn.v_experts.45", + "model.layers.9.self_attn.v_experts.46", + "model.layers.9.self_attn.v_experts.47", + "model.layers.9.self_attn.v_experts.48", + "model.layers.9.self_attn.v_experts.49", + "model.layers.9.self_attn.v_experts.50", + "model.layers.9.self_attn.v_experts.51", + "model.layers.9.self_attn.v_experts.52", + "model.layers.9.self_attn.v_experts.53", + "model.layers.9.self_attn.v_experts.54", + "model.layers.9.self_attn.v_experts.55", + "model.layers.9.self_attn.v_experts.56", + "model.layers.9.self_attn.v_experts.57", + "model.layers.9.self_attn.v_experts.58", + "model.layers.9.self_attn.v_experts.59", + "model.layers.9.self_attn.v_experts.60", + "model.layers.9.self_attn.v_experts.61", + "model.layers.9.self_attn.v_experts.62", + "model.layers.9.self_attn.v_experts.63", + "model.layers.9.self_attn.gate_proj", + "model.layers.9.mlp.gate", + "model.layers.9.mlp.shared_experts.gate_proj", + "model.layers.9.mlp.shared_experts.up_proj", + "model.layers.9.mlp.shared_experts.down_proj", + "model.layers.10.self_attn.q_proj", + "model.layers.10.self_attn.k_proj", + "model.layers.10.self_attn.o_proj", + "model.layers.10.self_attn.v_router", + "model.layers.10.self_attn.v_experts.0", + "model.layers.10.self_attn.v_experts.1", + "model.layers.10.self_attn.v_experts.2", + "model.layers.10.self_attn.v_experts.3", + "model.layers.10.self_attn.v_experts.4", + "model.layers.10.self_attn.v_experts.5", + "model.layers.10.self_attn.v_experts.6", + "model.layers.10.self_attn.v_experts.7", + "model.layers.10.self_attn.v_experts.8", + "model.layers.10.self_attn.v_experts.9", + "model.layers.10.self_attn.v_experts.10", + "model.layers.10.self_attn.v_experts.11", + "model.layers.10.self_attn.v_experts.12", + "model.layers.10.self_attn.v_experts.13", + "model.layers.10.self_attn.v_experts.14", + "model.layers.10.self_attn.v_experts.15", + "model.layers.10.self_attn.v_experts.16", + "model.layers.10.self_attn.v_experts.17", + "model.layers.10.self_attn.v_experts.18", + "model.layers.10.self_attn.v_experts.19", + "model.layers.10.self_attn.v_experts.20", + "model.layers.10.self_attn.v_experts.21", + "model.layers.10.self_attn.v_experts.22", + "model.layers.10.self_attn.v_experts.23", + "model.layers.10.self_attn.v_experts.24", + "model.layers.10.self_attn.v_experts.25", + "model.layers.10.self_attn.v_experts.26", + "model.layers.10.self_attn.v_experts.27", + "model.layers.10.self_attn.v_experts.28", + "model.layers.10.self_attn.v_experts.29", + "model.layers.10.self_attn.v_experts.30", + "model.layers.10.self_attn.v_experts.31", + "model.layers.10.self_attn.v_experts.32", + "model.layers.10.self_attn.v_experts.33", + "model.layers.10.self_attn.v_experts.34", + "model.layers.10.self_attn.v_experts.35", + "model.layers.10.self_attn.v_experts.36", + "model.layers.10.self_attn.v_experts.37", + "model.layers.10.self_attn.v_experts.38", + "model.layers.10.self_attn.v_experts.39", + "model.layers.10.self_attn.v_experts.40", + "model.layers.10.self_attn.v_experts.41", + "model.layers.10.self_attn.v_experts.42", + "model.layers.10.self_attn.v_experts.43", + "model.layers.10.self_attn.v_experts.44", + "model.layers.10.self_attn.v_experts.45", + "model.layers.10.self_attn.v_experts.46", + "model.layers.10.self_attn.v_experts.47", + "model.layers.10.self_attn.v_experts.48", + "model.layers.10.self_attn.v_experts.49", + "model.layers.10.self_attn.v_experts.50", + "model.layers.10.self_attn.v_experts.51", + "model.layers.10.self_attn.v_experts.52", + "model.layers.10.self_attn.v_experts.53", + "model.layers.10.self_attn.v_experts.54", + "model.layers.10.self_attn.v_experts.55", + "model.layers.10.self_attn.v_experts.56", + "model.layers.10.self_attn.v_experts.57", + "model.layers.10.self_attn.v_experts.58", + "model.layers.10.self_attn.v_experts.59", + "model.layers.10.self_attn.v_experts.60", + "model.layers.10.self_attn.v_experts.61", + "model.layers.10.self_attn.v_experts.62", + "model.layers.10.self_attn.v_experts.63", + "model.layers.10.self_attn.gate_proj", + "model.layers.10.mlp.gate", + "model.layers.10.mlp.shared_experts.gate_proj", + "model.layers.10.mlp.shared_experts.up_proj", + "model.layers.10.mlp.shared_experts.down_proj", + "model.layers.11.self_attn.q_proj", + "model.layers.11.self_attn.k_proj", + "model.layers.11.self_attn.o_proj", + "model.layers.11.self_attn.v_router", + "model.layers.11.self_attn.v_experts.0", + "model.layers.11.self_attn.v_experts.1", + "model.layers.11.self_attn.v_experts.2", + "model.layers.11.self_attn.v_experts.3", + "model.layers.11.self_attn.v_experts.4", + "model.layers.11.self_attn.v_experts.5", + "model.layers.11.self_attn.v_experts.6", + "model.layers.11.self_attn.v_experts.7", + "model.layers.11.self_attn.v_experts.8", + "model.layers.11.self_attn.v_experts.9", + "model.layers.11.self_attn.v_experts.10", + "model.layers.11.self_attn.v_experts.11", + "model.layers.11.self_attn.v_experts.12", + "model.layers.11.self_attn.v_experts.13", + "model.layers.11.self_attn.v_experts.14", + "model.layers.11.self_attn.v_experts.15", + "model.layers.11.self_attn.v_experts.16", + "model.layers.11.self_attn.v_experts.17", + "model.layers.11.self_attn.v_experts.18", + "model.layers.11.self_attn.v_experts.19", + "model.layers.11.self_attn.v_experts.20", + "model.layers.11.self_attn.v_experts.21", + "model.layers.11.self_attn.v_experts.22", + "model.layers.11.self_attn.v_experts.23", + "model.layers.11.self_attn.v_experts.24", + "model.layers.11.self_attn.v_experts.25", + "model.layers.11.self_attn.v_experts.26", + "model.layers.11.self_attn.v_experts.27", + "model.layers.11.self_attn.v_experts.28", + "model.layers.11.self_attn.v_experts.29", + "model.layers.11.self_attn.v_experts.30", + "model.layers.11.self_attn.v_experts.31", + "model.layers.11.self_attn.v_experts.32", + "model.layers.11.self_attn.v_experts.33", + "model.layers.11.self_attn.v_experts.34", + "model.layers.11.self_attn.v_experts.35", + "model.layers.11.self_attn.v_experts.36", + "model.layers.11.self_attn.v_experts.37", + "model.layers.11.self_attn.v_experts.38", + "model.layers.11.self_attn.v_experts.39", + "model.layers.11.self_attn.v_experts.40", + "model.layers.11.self_attn.v_experts.41", + "model.layers.11.self_attn.v_experts.42", + "model.layers.11.self_attn.v_experts.43", + "model.layers.11.self_attn.v_experts.44", + "model.layers.11.self_attn.v_experts.45", + "model.layers.11.self_attn.v_experts.46", + "model.layers.11.self_attn.v_experts.47", + "model.layers.11.self_attn.v_experts.48", + "model.layers.11.self_attn.v_experts.49", + "model.layers.11.self_attn.v_experts.50", + "model.layers.11.self_attn.v_experts.51", + "model.layers.11.self_attn.v_experts.52", + "model.layers.11.self_attn.v_experts.53", + "model.layers.11.self_attn.v_experts.54", + "model.layers.11.self_attn.v_experts.55", + "model.layers.11.self_attn.v_experts.56", + "model.layers.11.self_attn.v_experts.57", + "model.layers.11.self_attn.v_experts.58", + "model.layers.11.self_attn.v_experts.59", + "model.layers.11.self_attn.v_experts.60", + "model.layers.11.self_attn.v_experts.61", + "model.layers.11.self_attn.v_experts.62", + "model.layers.11.self_attn.v_experts.63", + "model.layers.11.self_attn.gate_proj", + "model.layers.11.mlp.gate", + "model.layers.11.mlp.shared_experts.gate_proj", + "model.layers.11.mlp.shared_experts.up_proj", + "model.layers.11.mlp.shared_experts.down_proj", + "model.layers.12.self_attn.q_proj", + "model.layers.12.self_attn.k_proj", + "model.layers.12.self_attn.o_proj", + "model.layers.12.self_attn.v_router", + "model.layers.12.self_attn.v_experts.0", + "model.layers.12.self_attn.v_experts.1", + "model.layers.12.self_attn.v_experts.2", + "model.layers.12.self_attn.v_experts.3", + "model.layers.12.self_attn.v_experts.4", + "model.layers.12.self_attn.v_experts.5", + "model.layers.12.self_attn.v_experts.6", + "model.layers.12.self_attn.v_experts.7", + "model.layers.12.self_attn.v_experts.8", + "model.layers.12.self_attn.v_experts.9", + "model.layers.12.self_attn.v_experts.10", + "model.layers.12.self_attn.v_experts.11", + "model.layers.12.self_attn.v_experts.12", + "model.layers.12.self_attn.v_experts.13", + "model.layers.12.self_attn.v_experts.14", + "model.layers.12.self_attn.v_experts.15", + "model.layers.12.self_attn.v_experts.16", + "model.layers.12.self_attn.v_experts.17", + "model.layers.12.self_attn.v_experts.18", + "model.layers.12.self_attn.v_experts.19", + "model.layers.12.self_attn.v_experts.20", + "model.layers.12.self_attn.v_experts.21", + "model.layers.12.self_attn.v_experts.22", + "model.layers.12.self_attn.v_experts.23", + "model.layers.12.self_attn.v_experts.24", + "model.layers.12.self_attn.v_experts.25", + "model.layers.12.self_attn.v_experts.26", + "model.layers.12.self_attn.v_experts.27", + "model.layers.12.self_attn.v_experts.28", + "model.layers.12.self_attn.v_experts.29", + "model.layers.12.self_attn.v_experts.30", + "model.layers.12.self_attn.v_experts.31", + "model.layers.12.self_attn.v_experts.32", + "model.layers.12.self_attn.v_experts.33", + "model.layers.12.self_attn.v_experts.34", + "model.layers.12.self_attn.v_experts.35", + "model.layers.12.self_attn.v_experts.36", + "model.layers.12.self_attn.v_experts.37", + "model.layers.12.self_attn.v_experts.38", + "model.layers.12.self_attn.v_experts.39", + "model.layers.12.self_attn.v_experts.40", + "model.layers.12.self_attn.v_experts.41", + "model.layers.12.self_attn.v_experts.42", + "model.layers.12.self_attn.v_experts.43", + "model.layers.12.self_attn.v_experts.44", + "model.layers.12.self_attn.v_experts.45", + "model.layers.12.self_attn.v_experts.46", + "model.layers.12.self_attn.v_experts.47", + "model.layers.12.self_attn.v_experts.48", + "model.layers.12.self_attn.v_experts.49", + "model.layers.12.self_attn.v_experts.50", + "model.layers.12.self_attn.v_experts.51", + "model.layers.12.self_attn.v_experts.52", + "model.layers.12.self_attn.v_experts.53", + "model.layers.12.self_attn.v_experts.54", + "model.layers.12.self_attn.v_experts.55", + "model.layers.12.self_attn.v_experts.56", + "model.layers.12.self_attn.v_experts.57", + "model.layers.12.self_attn.v_experts.58", + "model.layers.12.self_attn.v_experts.59", + "model.layers.12.self_attn.v_experts.60", + "model.layers.12.self_attn.v_experts.61", + "model.layers.12.self_attn.v_experts.62", + "model.layers.12.self_attn.v_experts.63", + "model.layers.12.self_attn.gate_proj", + "model.layers.12.mlp.gate", + "model.layers.12.mlp.shared_experts.gate_proj", + "model.layers.12.mlp.shared_experts.up_proj", + "model.layers.12.mlp.shared_experts.down_proj", + "model.layers.13.self_attn.q_proj", + "model.layers.13.self_attn.k_proj", + "model.layers.13.self_attn.o_proj", + "model.layers.13.self_attn.v_router", + "model.layers.13.self_attn.v_experts.0", + "model.layers.13.self_attn.v_experts.1", + "model.layers.13.self_attn.v_experts.2", + "model.layers.13.self_attn.v_experts.3", + "model.layers.13.self_attn.v_experts.4", + "model.layers.13.self_attn.v_experts.5", + "model.layers.13.self_attn.v_experts.6", + "model.layers.13.self_attn.v_experts.7", + "model.layers.13.self_attn.v_experts.8", + "model.layers.13.self_attn.v_experts.9", + "model.layers.13.self_attn.v_experts.10", + "model.layers.13.self_attn.v_experts.11", + "model.layers.13.self_attn.v_experts.12", + "model.layers.13.self_attn.v_experts.13", + "model.layers.13.self_attn.v_experts.14", + "model.layers.13.self_attn.v_experts.15", + "model.layers.13.self_attn.v_experts.16", + "model.layers.13.self_attn.v_experts.17", + "model.layers.13.self_attn.v_experts.18", + "model.layers.13.self_attn.v_experts.19", + "model.layers.13.self_attn.v_experts.20", + "model.layers.13.self_attn.v_experts.21", + "model.layers.13.self_attn.v_experts.22", + "model.layers.13.self_attn.v_experts.23", + "model.layers.13.self_attn.v_experts.24", + "model.layers.13.self_attn.v_experts.25", + "model.layers.13.self_attn.v_experts.26", + "model.layers.13.self_attn.v_experts.27", + "model.layers.13.self_attn.v_experts.28", + "model.layers.13.self_attn.v_experts.29", + "model.layers.13.self_attn.v_experts.30", + "model.layers.13.self_attn.v_experts.31", + "model.layers.13.self_attn.v_experts.32", + "model.layers.13.self_attn.v_experts.33", + "model.layers.13.self_attn.v_experts.34", + "model.layers.13.self_attn.v_experts.35", + "model.layers.13.self_attn.v_experts.36", + "model.layers.13.self_attn.v_experts.37", + "model.layers.13.self_attn.v_experts.38", + "model.layers.13.self_attn.v_experts.39", + "model.layers.13.self_attn.v_experts.40", + "model.layers.13.self_attn.v_experts.41", + "model.layers.13.self_attn.v_experts.42", + "model.layers.13.self_attn.v_experts.43", + "model.layers.13.self_attn.v_experts.44", + "model.layers.13.self_attn.v_experts.45", + "model.layers.13.self_attn.v_experts.46", + "model.layers.13.self_attn.v_experts.47", + "model.layers.13.self_attn.v_experts.48", + "model.layers.13.self_attn.v_experts.49", + "model.layers.13.self_attn.v_experts.50", + "model.layers.13.self_attn.v_experts.51", + "model.layers.13.self_attn.v_experts.52", + "model.layers.13.self_attn.v_experts.53", + "model.layers.13.self_attn.v_experts.54", + "model.layers.13.self_attn.v_experts.55", + "model.layers.13.self_attn.v_experts.56", + "model.layers.13.self_attn.v_experts.57", + "model.layers.13.self_attn.v_experts.58", + "model.layers.13.self_attn.v_experts.59", + "model.layers.13.self_attn.v_experts.60", + "model.layers.13.self_attn.v_experts.61", + "model.layers.13.self_attn.v_experts.62", + "model.layers.13.self_attn.v_experts.63", + "model.layers.13.self_attn.gate_proj", + "model.layers.13.mlp.gate", + "model.layers.13.mlp.shared_experts.gate_proj", + "model.layers.13.mlp.shared_experts.up_proj", + "model.layers.13.mlp.shared_experts.down_proj", + "model.layers.14.self_attn.q_proj", + "model.layers.14.self_attn.k_proj", + "model.layers.14.self_attn.o_proj", + "model.layers.14.self_attn.v_router", + "model.layers.14.self_attn.v_experts.0", + "model.layers.14.self_attn.v_experts.1", + "model.layers.14.self_attn.v_experts.2", + "model.layers.14.self_attn.v_experts.3", + "model.layers.14.self_attn.v_experts.4", + "model.layers.14.self_attn.v_experts.5", + "model.layers.14.self_attn.v_experts.6", + "model.layers.14.self_attn.v_experts.7", + "model.layers.14.self_attn.v_experts.8", + "model.layers.14.self_attn.v_experts.9", + "model.layers.14.self_attn.v_experts.10", + "model.layers.14.self_attn.v_experts.11", + "model.layers.14.self_attn.v_experts.12", + "model.layers.14.self_attn.v_experts.13", + "model.layers.14.self_attn.v_experts.14", + "model.layers.14.self_attn.v_experts.15", + "model.layers.14.self_attn.v_experts.16", + "model.layers.14.self_attn.v_experts.17", + "model.layers.14.self_attn.v_experts.18", + "model.layers.14.self_attn.v_experts.19", + "model.layers.14.self_attn.v_experts.20", + "model.layers.14.self_attn.v_experts.21", + "model.layers.14.self_attn.v_experts.22", + "model.layers.14.self_attn.v_experts.23", + "model.layers.14.self_attn.v_experts.24", + "model.layers.14.self_attn.v_experts.25", + "model.layers.14.self_attn.v_experts.26", + "model.layers.14.self_attn.v_experts.27", + "model.layers.14.self_attn.v_experts.28", + "model.layers.14.self_attn.v_experts.29", + "model.layers.14.self_attn.v_experts.30", + "model.layers.14.self_attn.v_experts.31", + "model.layers.14.self_attn.v_experts.32", + "model.layers.14.self_attn.v_experts.33", + "model.layers.14.self_attn.v_experts.34", + "model.layers.14.self_attn.v_experts.35", + "model.layers.14.self_attn.v_experts.36", + "model.layers.14.self_attn.v_experts.37", + "model.layers.14.self_attn.v_experts.38", + "model.layers.14.self_attn.v_experts.39", + "model.layers.14.self_attn.v_experts.40", + "model.layers.14.self_attn.v_experts.41", + "model.layers.14.self_attn.v_experts.42", + "model.layers.14.self_attn.v_experts.43", + "model.layers.14.self_attn.v_experts.44", + "model.layers.14.self_attn.v_experts.45", + "model.layers.14.self_attn.v_experts.46", + "model.layers.14.self_attn.v_experts.47", + "model.layers.14.self_attn.v_experts.48", + "model.layers.14.self_attn.v_experts.49", + "model.layers.14.self_attn.v_experts.50", + "model.layers.14.self_attn.v_experts.51", + "model.layers.14.self_attn.v_experts.52", + "model.layers.14.self_attn.v_experts.53", + "model.layers.14.self_attn.v_experts.54", + "model.layers.14.self_attn.v_experts.55", + "model.layers.14.self_attn.v_experts.56", + "model.layers.14.self_attn.v_experts.57", + "model.layers.14.self_attn.v_experts.58", + "model.layers.14.self_attn.v_experts.59", + "model.layers.14.self_attn.v_experts.60", + "model.layers.14.self_attn.v_experts.61", + "model.layers.14.self_attn.v_experts.62", + "model.layers.14.self_attn.v_experts.63", + "model.layers.14.self_attn.gate_proj", + "model.layers.14.mlp.gate", + "model.layers.14.mlp.shared_experts.gate_proj", + "model.layers.14.mlp.shared_experts.up_proj", + "model.layers.14.mlp.shared_experts.down_proj", + "model.layers.15.self_attn.q_proj", + "model.layers.15.self_attn.k_proj", + "model.layers.15.self_attn.o_proj", + "model.layers.15.self_attn.v_router", + "model.layers.15.self_attn.v_experts.0", + "model.layers.15.self_attn.v_experts.1", + "model.layers.15.self_attn.v_experts.2", + "model.layers.15.self_attn.v_experts.3", + "model.layers.15.self_attn.v_experts.4", + "model.layers.15.self_attn.v_experts.5", + "model.layers.15.self_attn.v_experts.6", + "model.layers.15.self_attn.v_experts.7", + "model.layers.15.self_attn.v_experts.8", + "model.layers.15.self_attn.v_experts.9", + "model.layers.15.self_attn.v_experts.10", + "model.layers.15.self_attn.v_experts.11", + "model.layers.15.self_attn.v_experts.12", + "model.layers.15.self_attn.v_experts.13", + "model.layers.15.self_attn.v_experts.14", + "model.layers.15.self_attn.v_experts.15", + "model.layers.15.self_attn.v_experts.16", + "model.layers.15.self_attn.v_experts.17", + "model.layers.15.self_attn.v_experts.18", + "model.layers.15.self_attn.v_experts.19", + "model.layers.15.self_attn.v_experts.20", + "model.layers.15.self_attn.v_experts.21", + "model.layers.15.self_attn.v_experts.22", + "model.layers.15.self_attn.v_experts.23", + "model.layers.15.self_attn.v_experts.24", + "model.layers.15.self_attn.v_experts.25", + "model.layers.15.self_attn.v_experts.26", + "model.layers.15.self_attn.v_experts.27", + "model.layers.15.self_attn.v_experts.28", + "model.layers.15.self_attn.v_experts.29", + "model.layers.15.self_attn.v_experts.30", + "model.layers.15.self_attn.v_experts.31", + "model.layers.15.self_attn.v_experts.32", + "model.layers.15.self_attn.v_experts.33", + "model.layers.15.self_attn.v_experts.34", + "model.layers.15.self_attn.v_experts.35", + "model.layers.15.self_attn.v_experts.36", + "model.layers.15.self_attn.v_experts.37", + "model.layers.15.self_attn.v_experts.38", + "model.layers.15.self_attn.v_experts.39", + "model.layers.15.self_attn.v_experts.40", + "model.layers.15.self_attn.v_experts.41", + "model.layers.15.self_attn.v_experts.42", + "model.layers.15.self_attn.v_experts.43", + "model.layers.15.self_attn.v_experts.44", + "model.layers.15.self_attn.v_experts.45", + "model.layers.15.self_attn.v_experts.46", + "model.layers.15.self_attn.v_experts.47", + "model.layers.15.self_attn.v_experts.48", + "model.layers.15.self_attn.v_experts.49", + "model.layers.15.self_attn.v_experts.50", + "model.layers.15.self_attn.v_experts.51", + "model.layers.15.self_attn.v_experts.52", + "model.layers.15.self_attn.v_experts.53", + "model.layers.15.self_attn.v_experts.54", + "model.layers.15.self_attn.v_experts.55", + "model.layers.15.self_attn.v_experts.56", + "model.layers.15.self_attn.v_experts.57", + "model.layers.15.self_attn.v_experts.58", + "model.layers.15.self_attn.v_experts.59", + "model.layers.15.self_attn.v_experts.60", + "model.layers.15.self_attn.v_experts.61", + "model.layers.15.self_attn.v_experts.62", + "model.layers.15.self_attn.v_experts.63", + "model.layers.15.self_attn.gate_proj", + "model.layers.15.mlp.gate", + "model.layers.15.mlp.shared_experts.gate_proj", + "model.layers.15.mlp.shared_experts.up_proj", + "model.layers.15.mlp.shared_experts.down_proj", + "model.layers.16.self_attn.q_proj", + "model.layers.16.self_attn.k_proj", + "model.layers.16.self_attn.o_proj", + "model.layers.16.self_attn.v_router", + "model.layers.16.self_attn.v_experts.0", + "model.layers.16.self_attn.v_experts.1", + "model.layers.16.self_attn.v_experts.2", + "model.layers.16.self_attn.v_experts.3", + "model.layers.16.self_attn.v_experts.4", + "model.layers.16.self_attn.v_experts.5", + "model.layers.16.self_attn.v_experts.6", + "model.layers.16.self_attn.v_experts.7", + "model.layers.16.self_attn.v_experts.8", + "model.layers.16.self_attn.v_experts.9", + "model.layers.16.self_attn.v_experts.10", + "model.layers.16.self_attn.v_experts.11", + "model.layers.16.self_attn.v_experts.12", + "model.layers.16.self_attn.v_experts.13", + "model.layers.16.self_attn.v_experts.14", + "model.layers.16.self_attn.v_experts.15", + "model.layers.16.self_attn.v_experts.16", + "model.layers.16.self_attn.v_experts.17", + "model.layers.16.self_attn.v_experts.18", + "model.layers.16.self_attn.v_experts.19", + "model.layers.16.self_attn.v_experts.20", + "model.layers.16.self_attn.v_experts.21", + "model.layers.16.self_attn.v_experts.22", + "model.layers.16.self_attn.v_experts.23", + "model.layers.16.self_attn.v_experts.24", + "model.layers.16.self_attn.v_experts.25", + "model.layers.16.self_attn.v_experts.26", + "model.layers.16.self_attn.v_experts.27", + "model.layers.16.self_attn.v_experts.28", + "model.layers.16.self_attn.v_experts.29", + "model.layers.16.self_attn.v_experts.30", + "model.layers.16.self_attn.v_experts.31", + "model.layers.16.self_attn.v_experts.32", + "model.layers.16.self_attn.v_experts.33", + "model.layers.16.self_attn.v_experts.34", + "model.layers.16.self_attn.v_experts.35", + "model.layers.16.self_attn.v_experts.36", + "model.layers.16.self_attn.v_experts.37", + "model.layers.16.self_attn.v_experts.38", + "model.layers.16.self_attn.v_experts.39", + "model.layers.16.self_attn.v_experts.40", + "model.layers.16.self_attn.v_experts.41", + "model.layers.16.self_attn.v_experts.42", + "model.layers.16.self_attn.v_experts.43", + "model.layers.16.self_attn.v_experts.44", + "model.layers.16.self_attn.v_experts.45", + "model.layers.16.self_attn.v_experts.46", + "model.layers.16.self_attn.v_experts.47", + "model.layers.16.self_attn.v_experts.48", + "model.layers.16.self_attn.v_experts.49", + "model.layers.16.self_attn.v_experts.50", + "model.layers.16.self_attn.v_experts.51", + "model.layers.16.self_attn.v_experts.52", + "model.layers.16.self_attn.v_experts.53", + "model.layers.16.self_attn.v_experts.54", + "model.layers.16.self_attn.v_experts.55", + "model.layers.16.self_attn.v_experts.56", + "model.layers.16.self_attn.v_experts.57", + "model.layers.16.self_attn.v_experts.58", + "model.layers.16.self_attn.v_experts.59", + "model.layers.16.self_attn.v_experts.60", + "model.layers.16.self_attn.v_experts.61", + "model.layers.16.self_attn.v_experts.62", + "model.layers.16.self_attn.v_experts.63", + "model.layers.16.self_attn.gate_proj", + "model.layers.16.mlp.gate", + "model.layers.16.mlp.shared_experts.gate_proj", + "model.layers.16.mlp.shared_experts.up_proj", + "model.layers.16.mlp.shared_experts.down_proj", + "model.layers.17.self_attn.q_proj", + "model.layers.17.self_attn.k_proj", + "model.layers.17.self_attn.o_proj", + "model.layers.17.self_attn.v_router", + "model.layers.17.self_attn.v_experts.0", + "model.layers.17.self_attn.v_experts.1", + "model.layers.17.self_attn.v_experts.2", + "model.layers.17.self_attn.v_experts.3", + "model.layers.17.self_attn.v_experts.4", + "model.layers.17.self_attn.v_experts.5", + "model.layers.17.self_attn.v_experts.6", + "model.layers.17.self_attn.v_experts.7", + "model.layers.17.self_attn.v_experts.8", + "model.layers.17.self_attn.v_experts.9", + "model.layers.17.self_attn.v_experts.10", + "model.layers.17.self_attn.v_experts.11", + "model.layers.17.self_attn.v_experts.12", + "model.layers.17.self_attn.v_experts.13", + "model.layers.17.self_attn.v_experts.14", + "model.layers.17.self_attn.v_experts.15", + "model.layers.17.self_attn.v_experts.16", + "model.layers.17.self_attn.v_experts.17", + "model.layers.17.self_attn.v_experts.18", + "model.layers.17.self_attn.v_experts.19", + "model.layers.17.self_attn.v_experts.20", + "model.layers.17.self_attn.v_experts.21", + "model.layers.17.self_attn.v_experts.22", + "model.layers.17.self_attn.v_experts.23", + "model.layers.17.self_attn.v_experts.24", + "model.layers.17.self_attn.v_experts.25", + "model.layers.17.self_attn.v_experts.26", + "model.layers.17.self_attn.v_experts.27", + "model.layers.17.self_attn.v_experts.28", + "model.layers.17.self_attn.v_experts.29", + "model.layers.17.self_attn.v_experts.30", + "model.layers.17.self_attn.v_experts.31", + "model.layers.17.self_attn.v_experts.32", + "model.layers.17.self_attn.v_experts.33", + "model.layers.17.self_attn.v_experts.34", + "model.layers.17.self_attn.v_experts.35", + "model.layers.17.self_attn.v_experts.36", + "model.layers.17.self_attn.v_experts.37", + "model.layers.17.self_attn.v_experts.38", + "model.layers.17.self_attn.v_experts.39", + "model.layers.17.self_attn.v_experts.40", + "model.layers.17.self_attn.v_experts.41", + "model.layers.17.self_attn.v_experts.42", + "model.layers.17.self_attn.v_experts.43", + "model.layers.17.self_attn.v_experts.44", + "model.layers.17.self_attn.v_experts.45", + "model.layers.17.self_attn.v_experts.46", + "model.layers.17.self_attn.v_experts.47", + "model.layers.17.self_attn.v_experts.48", + "model.layers.17.self_attn.v_experts.49", + "model.layers.17.self_attn.v_experts.50", + "model.layers.17.self_attn.v_experts.51", + "model.layers.17.self_attn.v_experts.52", + "model.layers.17.self_attn.v_experts.53", + "model.layers.17.self_attn.v_experts.54", + "model.layers.17.self_attn.v_experts.55", + "model.layers.17.self_attn.v_experts.56", + "model.layers.17.self_attn.v_experts.57", + "model.layers.17.self_attn.v_experts.58", + "model.layers.17.self_attn.v_experts.59", + "model.layers.17.self_attn.v_experts.60", + "model.layers.17.self_attn.v_experts.61", + "model.layers.17.self_attn.v_experts.62", + "model.layers.17.self_attn.v_experts.63", + "model.layers.17.self_attn.gate_proj", + "model.layers.17.mlp.gate", + "model.layers.17.mlp.shared_experts.gate_proj", + "model.layers.17.mlp.shared_experts.up_proj", + "model.layers.17.mlp.shared_experts.down_proj", + "model.layers.18.self_attn.q_proj", + "model.layers.18.self_attn.k_proj", + "model.layers.18.self_attn.o_proj", + "model.layers.18.self_attn.v_router", + "model.layers.18.self_attn.v_experts.0", + "model.layers.18.self_attn.v_experts.1", + "model.layers.18.self_attn.v_experts.2", + "model.layers.18.self_attn.v_experts.3", + "model.layers.18.self_attn.v_experts.4", + "model.layers.18.self_attn.v_experts.5", + "model.layers.18.self_attn.v_experts.6", + "model.layers.18.self_attn.v_experts.7", + "model.layers.18.self_attn.v_experts.8", + "model.layers.18.self_attn.v_experts.9", + "model.layers.18.self_attn.v_experts.10", + "model.layers.18.self_attn.v_experts.11", + "model.layers.18.self_attn.v_experts.12", + "model.layers.18.self_attn.v_experts.13", + "model.layers.18.self_attn.v_experts.14", + "model.layers.18.self_attn.v_experts.15", + "model.layers.18.self_attn.v_experts.16", + "model.layers.18.self_attn.v_experts.17", + "model.layers.18.self_attn.v_experts.18", + "model.layers.18.self_attn.v_experts.19", + "model.layers.18.self_attn.v_experts.20", + "model.layers.18.self_attn.v_experts.21", + "model.layers.18.self_attn.v_experts.22", + "model.layers.18.self_attn.v_experts.23", + "model.layers.18.self_attn.v_experts.24", + "model.layers.18.self_attn.v_experts.25", + "model.layers.18.self_attn.v_experts.26", + "model.layers.18.self_attn.v_experts.27", + "model.layers.18.self_attn.v_experts.28", + "model.layers.18.self_attn.v_experts.29", + "model.layers.18.self_attn.v_experts.30", + "model.layers.18.self_attn.v_experts.31", + "model.layers.18.self_attn.v_experts.32", + "model.layers.18.self_attn.v_experts.33", + "model.layers.18.self_attn.v_experts.34", + "model.layers.18.self_attn.v_experts.35", + "model.layers.18.self_attn.v_experts.36", + "model.layers.18.self_attn.v_experts.37", + "model.layers.18.self_attn.v_experts.38", + "model.layers.18.self_attn.v_experts.39", + "model.layers.18.self_attn.v_experts.40", + "model.layers.18.self_attn.v_experts.41", + "model.layers.18.self_attn.v_experts.42", + "model.layers.18.self_attn.v_experts.43", + "model.layers.18.self_attn.v_experts.44", + "model.layers.18.self_attn.v_experts.45", + "model.layers.18.self_attn.v_experts.46", + "model.layers.18.self_attn.v_experts.47", + "model.layers.18.self_attn.v_experts.48", + "model.layers.18.self_attn.v_experts.49", + "model.layers.18.self_attn.v_experts.50", + "model.layers.18.self_attn.v_experts.51", + "model.layers.18.self_attn.v_experts.52", + "model.layers.18.self_attn.v_experts.53", + "model.layers.18.self_attn.v_experts.54", + "model.layers.18.self_attn.v_experts.55", + "model.layers.18.self_attn.v_experts.56", + "model.layers.18.self_attn.v_experts.57", + "model.layers.18.self_attn.v_experts.58", + "model.layers.18.self_attn.v_experts.59", + "model.layers.18.self_attn.v_experts.60", + "model.layers.18.self_attn.v_experts.61", + "model.layers.18.self_attn.v_experts.62", + "model.layers.18.self_attn.v_experts.63", + "model.layers.18.self_attn.gate_proj", + "model.layers.18.mlp.gate", + "model.layers.18.mlp.shared_experts.gate_proj", + "model.layers.18.mlp.shared_experts.up_proj", + "model.layers.18.mlp.shared_experts.down_proj", + "model.layers.19.self_attn.q_proj", + "model.layers.19.self_attn.k_proj", + "model.layers.19.self_attn.o_proj", + "model.layers.19.self_attn.v_router", + "model.layers.19.self_attn.v_experts.0", + "model.layers.19.self_attn.v_experts.1", + "model.layers.19.self_attn.v_experts.2", + "model.layers.19.self_attn.v_experts.3", + "model.layers.19.self_attn.v_experts.4", + "model.layers.19.self_attn.v_experts.5", + "model.layers.19.self_attn.v_experts.6", + "model.layers.19.self_attn.v_experts.7", + "model.layers.19.self_attn.v_experts.8", + "model.layers.19.self_attn.v_experts.9", + "model.layers.19.self_attn.v_experts.10", + "model.layers.19.self_attn.v_experts.11", + "model.layers.19.self_attn.v_experts.12", + "model.layers.19.self_attn.v_experts.13", + "model.layers.19.self_attn.v_experts.14", + "model.layers.19.self_attn.v_experts.15", + "model.layers.19.self_attn.v_experts.16", + "model.layers.19.self_attn.v_experts.17", + "model.layers.19.self_attn.v_experts.18", + "model.layers.19.self_attn.v_experts.19", + "model.layers.19.self_attn.v_experts.20", + "model.layers.19.self_attn.v_experts.21", + "model.layers.19.self_attn.v_experts.22", + "model.layers.19.self_attn.v_experts.23", + "model.layers.19.self_attn.v_experts.24", + "model.layers.19.self_attn.v_experts.25", + "model.layers.19.self_attn.v_experts.26", + "model.layers.19.self_attn.v_experts.27", + "model.layers.19.self_attn.v_experts.28", + "model.layers.19.self_attn.v_experts.29", + "model.layers.19.self_attn.v_experts.30", + "model.layers.19.self_attn.v_experts.31", + "model.layers.19.self_attn.v_experts.32", + "model.layers.19.self_attn.v_experts.33", + "model.layers.19.self_attn.v_experts.34", + "model.layers.19.self_attn.v_experts.35", + "model.layers.19.self_attn.v_experts.36", + "model.layers.19.self_attn.v_experts.37", + "model.layers.19.self_attn.v_experts.38", + "model.layers.19.self_attn.v_experts.39", + "model.layers.19.self_attn.v_experts.40", + "model.layers.19.self_attn.v_experts.41", + "model.layers.19.self_attn.v_experts.42", + "model.layers.19.self_attn.v_experts.43", + "model.layers.19.self_attn.v_experts.44", + "model.layers.19.self_attn.v_experts.45", + "model.layers.19.self_attn.v_experts.46", + "model.layers.19.self_attn.v_experts.47", + "model.layers.19.self_attn.v_experts.48", + "model.layers.19.self_attn.v_experts.49", + "model.layers.19.self_attn.v_experts.50", + "model.layers.19.self_attn.v_experts.51", + "model.layers.19.self_attn.v_experts.52", + "model.layers.19.self_attn.v_experts.53", + "model.layers.19.self_attn.v_experts.54", + "model.layers.19.self_attn.v_experts.55", + "model.layers.19.self_attn.v_experts.56", + "model.layers.19.self_attn.v_experts.57", + "model.layers.19.self_attn.v_experts.58", + "model.layers.19.self_attn.v_experts.59", + "model.layers.19.self_attn.v_experts.60", + "model.layers.19.self_attn.v_experts.61", + "model.layers.19.self_attn.v_experts.62", + "model.layers.19.self_attn.v_experts.63", + "model.layers.19.self_attn.gate_proj", + "model.layers.19.mlp.gate", + "model.layers.19.mlp.shared_experts.gate_proj", + "model.layers.19.mlp.shared_experts.up_proj", + "model.layers.19.mlp.shared_experts.down_proj", + "model.layers.20.self_attn.q_proj", + "model.layers.20.self_attn.k_proj", + "model.layers.20.self_attn.o_proj", + "model.layers.20.self_attn.v_router", + "model.layers.20.self_attn.v_experts.0", + "model.layers.20.self_attn.v_experts.1", + "model.layers.20.self_attn.v_experts.2", + "model.layers.20.self_attn.v_experts.3", + "model.layers.20.self_attn.v_experts.4", + "model.layers.20.self_attn.v_experts.5", + "model.layers.20.self_attn.v_experts.6", + "model.layers.20.self_attn.v_experts.7", + "model.layers.20.self_attn.v_experts.8", + "model.layers.20.self_attn.v_experts.9", + "model.layers.20.self_attn.v_experts.10", + "model.layers.20.self_attn.v_experts.11", + "model.layers.20.self_attn.v_experts.12", + "model.layers.20.self_attn.v_experts.13", + "model.layers.20.self_attn.v_experts.14", + "model.layers.20.self_attn.v_experts.15", + "model.layers.20.self_attn.v_experts.16", + "model.layers.20.self_attn.v_experts.17", + "model.layers.20.self_attn.v_experts.18", + "model.layers.20.self_attn.v_experts.19", + "model.layers.20.self_attn.v_experts.20", + "model.layers.20.self_attn.v_experts.21", + "model.layers.20.self_attn.v_experts.22", + "model.layers.20.self_attn.v_experts.23", + "model.layers.20.self_attn.v_experts.24", + "model.layers.20.self_attn.v_experts.25", + "model.layers.20.self_attn.v_experts.26", + "model.layers.20.self_attn.v_experts.27", + "model.layers.20.self_attn.v_experts.28", + "model.layers.20.self_attn.v_experts.29", + "model.layers.20.self_attn.v_experts.30", + "model.layers.20.self_attn.v_experts.31", + "model.layers.20.self_attn.v_experts.32", + "model.layers.20.self_attn.v_experts.33", + "model.layers.20.self_attn.v_experts.34", + "model.layers.20.self_attn.v_experts.35", + "model.layers.20.self_attn.v_experts.36", + "model.layers.20.self_attn.v_experts.37", + "model.layers.20.self_attn.v_experts.38", + "model.layers.20.self_attn.v_experts.39", + "model.layers.20.self_attn.v_experts.40", + "model.layers.20.self_attn.v_experts.41", + "model.layers.20.self_attn.v_experts.42", + "model.layers.20.self_attn.v_experts.43", + "model.layers.20.self_attn.v_experts.44", + "model.layers.20.self_attn.v_experts.45", + "model.layers.20.self_attn.v_experts.46", + "model.layers.20.self_attn.v_experts.47", + "model.layers.20.self_attn.v_experts.48", + "model.layers.20.self_attn.v_experts.49", + "model.layers.20.self_attn.v_experts.50", + "model.layers.20.self_attn.v_experts.51", + "model.layers.20.self_attn.v_experts.52", + "model.layers.20.self_attn.v_experts.53", + "model.layers.20.self_attn.v_experts.54", + "model.layers.20.self_attn.v_experts.55", + "model.layers.20.self_attn.v_experts.56", + "model.layers.20.self_attn.v_experts.57", + "model.layers.20.self_attn.v_experts.58", + "model.layers.20.self_attn.v_experts.59", + "model.layers.20.self_attn.v_experts.60", + "model.layers.20.self_attn.v_experts.61", + "model.layers.20.self_attn.v_experts.62", + "model.layers.20.self_attn.v_experts.63", + "model.layers.20.self_attn.gate_proj", + "model.layers.20.mlp.gate", + "model.layers.20.mlp.shared_experts.gate_proj", + "model.layers.20.mlp.shared_experts.up_proj", + "model.layers.20.mlp.shared_experts.down_proj", + "model.layers.21.self_attn.q_proj", + "model.layers.21.self_attn.k_proj", + "model.layers.21.self_attn.o_proj", + "model.layers.21.self_attn.v_router", + "model.layers.21.self_attn.v_experts.0", + "model.layers.21.self_attn.v_experts.1", + "model.layers.21.self_attn.v_experts.2", + "model.layers.21.self_attn.v_experts.3", + "model.layers.21.self_attn.v_experts.4", + "model.layers.21.self_attn.v_experts.5", + "model.layers.21.self_attn.v_experts.6", + "model.layers.21.self_attn.v_experts.7", + "model.layers.21.self_attn.v_experts.8", + "model.layers.21.self_attn.v_experts.9", + "model.layers.21.self_attn.v_experts.10", + "model.layers.21.self_attn.v_experts.11", + "model.layers.21.self_attn.v_experts.12", + "model.layers.21.self_attn.v_experts.13", + "model.layers.21.self_attn.v_experts.14", + "model.layers.21.self_attn.v_experts.15", + "model.layers.21.self_attn.v_experts.16", + "model.layers.21.self_attn.v_experts.17", + "model.layers.21.self_attn.v_experts.18", + "model.layers.21.self_attn.v_experts.19", + "model.layers.21.self_attn.v_experts.20", + "model.layers.21.self_attn.v_experts.21", + "model.layers.21.self_attn.v_experts.22", + "model.layers.21.self_attn.v_experts.23", + "model.layers.21.self_attn.v_experts.24", + "model.layers.21.self_attn.v_experts.25", + "model.layers.21.self_attn.v_experts.26", + "model.layers.21.self_attn.v_experts.27", + "model.layers.21.self_attn.v_experts.28", + "model.layers.21.self_attn.v_experts.29", + "model.layers.21.self_attn.v_experts.30", + "model.layers.21.self_attn.v_experts.31", + "model.layers.21.self_attn.v_experts.32", + "model.layers.21.self_attn.v_experts.33", + "model.layers.21.self_attn.v_experts.34", + "model.layers.21.self_attn.v_experts.35", + "model.layers.21.self_attn.v_experts.36", + "model.layers.21.self_attn.v_experts.37", + "model.layers.21.self_attn.v_experts.38", + "model.layers.21.self_attn.v_experts.39", + "model.layers.21.self_attn.v_experts.40", + "model.layers.21.self_attn.v_experts.41", + "model.layers.21.self_attn.v_experts.42", + "model.layers.21.self_attn.v_experts.43", + "model.layers.21.self_attn.v_experts.44", + "model.layers.21.self_attn.v_experts.45", + "model.layers.21.self_attn.v_experts.46", + "model.layers.21.self_attn.v_experts.47", + "model.layers.21.self_attn.v_experts.48", + "model.layers.21.self_attn.v_experts.49", + "model.layers.21.self_attn.v_experts.50", + "model.layers.21.self_attn.v_experts.51", + "model.layers.21.self_attn.v_experts.52", + "model.layers.21.self_attn.v_experts.53", + "model.layers.21.self_attn.v_experts.54", + "model.layers.21.self_attn.v_experts.55", + "model.layers.21.self_attn.v_experts.56", + "model.layers.21.self_attn.v_experts.57", + "model.layers.21.self_attn.v_experts.58", + "model.layers.21.self_attn.v_experts.59", + "model.layers.21.self_attn.v_experts.60", + "model.layers.21.self_attn.v_experts.61", + "model.layers.21.self_attn.v_experts.62", + "model.layers.21.self_attn.v_experts.63", + "model.layers.21.self_attn.gate_proj", + "model.layers.21.mlp.gate", + "model.layers.21.mlp.shared_experts.gate_proj", + "model.layers.21.mlp.shared_experts.up_proj", + "model.layers.21.mlp.shared_experts.down_proj", + "model.layers.22.self_attn.q_proj", + "model.layers.22.self_attn.k_proj", + "model.layers.22.self_attn.o_proj", + "model.layers.22.self_attn.v_router", + "model.layers.22.self_attn.v_experts.0", + "model.layers.22.self_attn.v_experts.1", + "model.layers.22.self_attn.v_experts.2", + "model.layers.22.self_attn.v_experts.3", + "model.layers.22.self_attn.v_experts.4", + "model.layers.22.self_attn.v_experts.5", + "model.layers.22.self_attn.v_experts.6", + "model.layers.22.self_attn.v_experts.7", + "model.layers.22.self_attn.v_experts.8", + "model.layers.22.self_attn.v_experts.9", + "model.layers.22.self_attn.v_experts.10", + "model.layers.22.self_attn.v_experts.11", + "model.layers.22.self_attn.v_experts.12", + "model.layers.22.self_attn.v_experts.13", + "model.layers.22.self_attn.v_experts.14", + "model.layers.22.self_attn.v_experts.15", + "model.layers.22.self_attn.v_experts.16", + "model.layers.22.self_attn.v_experts.17", + "model.layers.22.self_attn.v_experts.18", + "model.layers.22.self_attn.v_experts.19", + "model.layers.22.self_attn.v_experts.20", + "model.layers.22.self_attn.v_experts.21", + "model.layers.22.self_attn.v_experts.22", + "model.layers.22.self_attn.v_experts.23", + "model.layers.22.self_attn.v_experts.24", + "model.layers.22.self_attn.v_experts.25", + "model.layers.22.self_attn.v_experts.26", + "model.layers.22.self_attn.v_experts.27", + "model.layers.22.self_attn.v_experts.28", + "model.layers.22.self_attn.v_experts.29", + "model.layers.22.self_attn.v_experts.30", + "model.layers.22.self_attn.v_experts.31", + "model.layers.22.self_attn.v_experts.32", + "model.layers.22.self_attn.v_experts.33", + "model.layers.22.self_attn.v_experts.34", + "model.layers.22.self_attn.v_experts.35", + "model.layers.22.self_attn.v_experts.36", + "model.layers.22.self_attn.v_experts.37", + "model.layers.22.self_attn.v_experts.38", + "model.layers.22.self_attn.v_experts.39", + "model.layers.22.self_attn.v_experts.40", + "model.layers.22.self_attn.v_experts.41", + "model.layers.22.self_attn.v_experts.42", + "model.layers.22.self_attn.v_experts.43", + "model.layers.22.self_attn.v_experts.44", + "model.layers.22.self_attn.v_experts.45", + "model.layers.22.self_attn.v_experts.46", + "model.layers.22.self_attn.v_experts.47", + "model.layers.22.self_attn.v_experts.48", + "model.layers.22.self_attn.v_experts.49", + "model.layers.22.self_attn.v_experts.50", + "model.layers.22.self_attn.v_experts.51", + "model.layers.22.self_attn.v_experts.52", + "model.layers.22.self_attn.v_experts.53", + "model.layers.22.self_attn.v_experts.54", + "model.layers.22.self_attn.v_experts.55", + "model.layers.22.self_attn.v_experts.56", + "model.layers.22.self_attn.v_experts.57", + "model.layers.22.self_attn.v_experts.58", + "model.layers.22.self_attn.v_experts.59", + "model.layers.22.self_attn.v_experts.60", + "model.layers.22.self_attn.v_experts.61", + "model.layers.22.self_attn.v_experts.62", + "model.layers.22.self_attn.v_experts.63", + "model.layers.22.self_attn.gate_proj", + "model.layers.22.mlp.gate", + "model.layers.22.mlp.shared_experts.gate_proj", + "model.layers.22.mlp.shared_experts.up_proj", + "model.layers.22.mlp.shared_experts.down_proj", + "model.layers.23.self_attn.q_proj", + "model.layers.23.self_attn.k_proj", + "model.layers.23.self_attn.o_proj", + "model.layers.23.self_attn.v_router", + "model.layers.23.self_attn.v_experts.0", + "model.layers.23.self_attn.v_experts.1", + "model.layers.23.self_attn.v_experts.2", + "model.layers.23.self_attn.v_experts.3", + "model.layers.23.self_attn.v_experts.4", + "model.layers.23.self_attn.v_experts.5", + "model.layers.23.self_attn.v_experts.6", + "model.layers.23.self_attn.v_experts.7", + "model.layers.23.self_attn.v_experts.8", + "model.layers.23.self_attn.v_experts.9", + "model.layers.23.self_attn.v_experts.10", + "model.layers.23.self_attn.v_experts.11", + "model.layers.23.self_attn.v_experts.12", + "model.layers.23.self_attn.v_experts.13", + "model.layers.23.self_attn.v_experts.14", + "model.layers.23.self_attn.v_experts.15", + "model.layers.23.self_attn.v_experts.16", + "model.layers.23.self_attn.v_experts.17", + "model.layers.23.self_attn.v_experts.18", + "model.layers.23.self_attn.v_experts.19", + "model.layers.23.self_attn.v_experts.20", + "model.layers.23.self_attn.v_experts.21", + "model.layers.23.self_attn.v_experts.22", + "model.layers.23.self_attn.v_experts.23", + "model.layers.23.self_attn.v_experts.24", + "model.layers.23.self_attn.v_experts.25", + "model.layers.23.self_attn.v_experts.26", + "model.layers.23.self_attn.v_experts.27", + "model.layers.23.self_attn.v_experts.28", + "model.layers.23.self_attn.v_experts.29", + "model.layers.23.self_attn.v_experts.30", + "model.layers.23.self_attn.v_experts.31", + "model.layers.23.self_attn.v_experts.32", + "model.layers.23.self_attn.v_experts.33", + "model.layers.23.self_attn.v_experts.34", + "model.layers.23.self_attn.v_experts.35", + "model.layers.23.self_attn.v_experts.36", + "model.layers.23.self_attn.v_experts.37", + "model.layers.23.self_attn.v_experts.38", + "model.layers.23.self_attn.v_experts.39", + "model.layers.23.self_attn.v_experts.40", + "model.layers.23.self_attn.v_experts.41", + "model.layers.23.self_attn.v_experts.42", + "model.layers.23.self_attn.v_experts.43", + "model.layers.23.self_attn.v_experts.44", + "model.layers.23.self_attn.v_experts.45", + "model.layers.23.self_attn.v_experts.46", + "model.layers.23.self_attn.v_experts.47", + "model.layers.23.self_attn.v_experts.48", + "model.layers.23.self_attn.v_experts.49", + "model.layers.23.self_attn.v_experts.50", + "model.layers.23.self_attn.v_experts.51", + "model.layers.23.self_attn.v_experts.52", + "model.layers.23.self_attn.v_experts.53", + "model.layers.23.self_attn.v_experts.54", + "model.layers.23.self_attn.v_experts.55", + "model.layers.23.self_attn.v_experts.56", + "model.layers.23.self_attn.v_experts.57", + "model.layers.23.self_attn.v_experts.58", + "model.layers.23.self_attn.v_experts.59", + "model.layers.23.self_attn.v_experts.60", + "model.layers.23.self_attn.v_experts.61", + "model.layers.23.self_attn.v_experts.62", + "model.layers.23.self_attn.v_experts.63", + "model.layers.23.self_attn.gate_proj", + "model.layers.23.mlp.gate", + "model.layers.23.mlp.shared_experts.gate_proj", + "model.layers.23.mlp.shared_experts.up_proj", + "model.layers.23.mlp.shared_experts.down_proj", + "model.layers.24.self_attn.q_proj", + "model.layers.24.self_attn.k_proj", + "model.layers.24.self_attn.o_proj", + "model.layers.24.self_attn.v_router", + "model.layers.24.self_attn.v_experts.0", + "model.layers.24.self_attn.v_experts.1", + "model.layers.24.self_attn.v_experts.2", + "model.layers.24.self_attn.v_experts.3", + "model.layers.24.self_attn.v_experts.4", + "model.layers.24.self_attn.v_experts.5", + "model.layers.24.self_attn.v_experts.6", + "model.layers.24.self_attn.v_experts.7", + "model.layers.24.self_attn.v_experts.8", + "model.layers.24.self_attn.v_experts.9", + "model.layers.24.self_attn.v_experts.10", + "model.layers.24.self_attn.v_experts.11", + "model.layers.24.self_attn.v_experts.12", + "model.layers.24.self_attn.v_experts.13", + "model.layers.24.self_attn.v_experts.14", + "model.layers.24.self_attn.v_experts.15", + "model.layers.24.self_attn.v_experts.16", + "model.layers.24.self_attn.v_experts.17", + "model.layers.24.self_attn.v_experts.18", + "model.layers.24.self_attn.v_experts.19", + "model.layers.24.self_attn.v_experts.20", + "model.layers.24.self_attn.v_experts.21", + "model.layers.24.self_attn.v_experts.22", + "model.layers.24.self_attn.v_experts.23", + "model.layers.24.self_attn.v_experts.24", + "model.layers.24.self_attn.v_experts.25", + "model.layers.24.self_attn.v_experts.26", + "model.layers.24.self_attn.v_experts.27", + "model.layers.24.self_attn.v_experts.28", + "model.layers.24.self_attn.v_experts.29", + "model.layers.24.self_attn.v_experts.30", + "model.layers.24.self_attn.v_experts.31", + "model.layers.24.self_attn.v_experts.32", + "model.layers.24.self_attn.v_experts.33", + "model.layers.24.self_attn.v_experts.34", + "model.layers.24.self_attn.v_experts.35", + "model.layers.24.self_attn.v_experts.36", + "model.layers.24.self_attn.v_experts.37", + "model.layers.24.self_attn.v_experts.38", + "model.layers.24.self_attn.v_experts.39", + "model.layers.24.self_attn.v_experts.40", + "model.layers.24.self_attn.v_experts.41", + "model.layers.24.self_attn.v_experts.42", + "model.layers.24.self_attn.v_experts.43", + "model.layers.24.self_attn.v_experts.44", + "model.layers.24.self_attn.v_experts.45", + "model.layers.24.self_attn.v_experts.46", + "model.layers.24.self_attn.v_experts.47", + "model.layers.24.self_attn.v_experts.48", + "model.layers.24.self_attn.v_experts.49", + "model.layers.24.self_attn.v_experts.50", + "model.layers.24.self_attn.v_experts.51", + "model.layers.24.self_attn.v_experts.52", + "model.layers.24.self_attn.v_experts.53", + "model.layers.24.self_attn.v_experts.54", + "model.layers.24.self_attn.v_experts.55", + "model.layers.24.self_attn.v_experts.56", + "model.layers.24.self_attn.v_experts.57", + "model.layers.24.self_attn.v_experts.58", + "model.layers.24.self_attn.v_experts.59", + "model.layers.24.self_attn.v_experts.60", + "model.layers.24.self_attn.v_experts.61", + "model.layers.24.self_attn.v_experts.62", + "model.layers.24.self_attn.v_experts.63", + "model.layers.24.self_attn.gate_proj", + "model.layers.24.mlp.gate", + "model.layers.24.mlp.shared_experts.gate_proj", + "model.layers.24.mlp.shared_experts.up_proj", + "model.layers.24.mlp.shared_experts.down_proj", + "model.layers.25.self_attn.q_proj", + "model.layers.25.self_attn.k_proj", + "model.layers.25.self_attn.o_proj", + "model.layers.25.self_attn.v_router", + "model.layers.25.self_attn.v_experts.0", + "model.layers.25.self_attn.v_experts.1", + "model.layers.25.self_attn.v_experts.2", + "model.layers.25.self_attn.v_experts.3", + "model.layers.25.self_attn.v_experts.4", + "model.layers.25.self_attn.v_experts.5", + "model.layers.25.self_attn.v_experts.6", + "model.layers.25.self_attn.v_experts.7", + "model.layers.25.self_attn.v_experts.8", + "model.layers.25.self_attn.v_experts.9", + "model.layers.25.self_attn.v_experts.10", + "model.layers.25.self_attn.v_experts.11", + "model.layers.25.self_attn.v_experts.12", + "model.layers.25.self_attn.v_experts.13", + "model.layers.25.self_attn.v_experts.14", + "model.layers.25.self_attn.v_experts.15", + "model.layers.25.self_attn.v_experts.16", + "model.layers.25.self_attn.v_experts.17", + "model.layers.25.self_attn.v_experts.18", + "model.layers.25.self_attn.v_experts.19", + "model.layers.25.self_attn.v_experts.20", + "model.layers.25.self_attn.v_experts.21", + "model.layers.25.self_attn.v_experts.22", + "model.layers.25.self_attn.v_experts.23", + "model.layers.25.self_attn.v_experts.24", + "model.layers.25.self_attn.v_experts.25", + "model.layers.25.self_attn.v_experts.26", + "model.layers.25.self_attn.v_experts.27", + "model.layers.25.self_attn.v_experts.28", + "model.layers.25.self_attn.v_experts.29", + "model.layers.25.self_attn.v_experts.30", + "model.layers.25.self_attn.v_experts.31", + "model.layers.25.self_attn.v_experts.32", + "model.layers.25.self_attn.v_experts.33", + "model.layers.25.self_attn.v_experts.34", + "model.layers.25.self_attn.v_experts.35", + "model.layers.25.self_attn.v_experts.36", + "model.layers.25.self_attn.v_experts.37", + "model.layers.25.self_attn.v_experts.38", + "model.layers.25.self_attn.v_experts.39", + "model.layers.25.self_attn.v_experts.40", + "model.layers.25.self_attn.v_experts.41", + "model.layers.25.self_attn.v_experts.42", + "model.layers.25.self_attn.v_experts.43", + "model.layers.25.self_attn.v_experts.44", + "model.layers.25.self_attn.v_experts.45", + "model.layers.25.self_attn.v_experts.46", + "model.layers.25.self_attn.v_experts.47", + "model.layers.25.self_attn.v_experts.48", + "model.layers.25.self_attn.v_experts.49", + "model.layers.25.self_attn.v_experts.50", + "model.layers.25.self_attn.v_experts.51", + "model.layers.25.self_attn.v_experts.52", + "model.layers.25.self_attn.v_experts.53", + "model.layers.25.self_attn.v_experts.54", + "model.layers.25.self_attn.v_experts.55", + "model.layers.25.self_attn.v_experts.56", + "model.layers.25.self_attn.v_experts.57", + "model.layers.25.self_attn.v_experts.58", + "model.layers.25.self_attn.v_experts.59", + "model.layers.25.self_attn.v_experts.60", + "model.layers.25.self_attn.v_experts.61", + "model.layers.25.self_attn.v_experts.62", + "model.layers.25.self_attn.v_experts.63", + "model.layers.25.self_attn.gate_proj", + "model.layers.25.mlp.gate", + "model.layers.25.mlp.shared_experts.gate_proj", + "model.layers.25.mlp.shared_experts.up_proj", + "model.layers.25.mlp.shared_experts.down_proj", + "model.layers.26.self_attn.q_proj", + "model.layers.26.self_attn.k_proj", + "model.layers.26.self_attn.o_proj", + "model.layers.26.self_attn.v_router", + "model.layers.26.self_attn.v_experts.0", + "model.layers.26.self_attn.v_experts.1", + "model.layers.26.self_attn.v_experts.2", + "model.layers.26.self_attn.v_experts.3", + "model.layers.26.self_attn.v_experts.4", + "model.layers.26.self_attn.v_experts.5", + "model.layers.26.self_attn.v_experts.6", + "model.layers.26.self_attn.v_experts.7", + "model.layers.26.self_attn.v_experts.8", + "model.layers.26.self_attn.v_experts.9", + "model.layers.26.self_attn.v_experts.10", + "model.layers.26.self_attn.v_experts.11", + "model.layers.26.self_attn.v_experts.12", + "model.layers.26.self_attn.v_experts.13", + "model.layers.26.self_attn.v_experts.14", + "model.layers.26.self_attn.v_experts.15", + "model.layers.26.self_attn.v_experts.16", + "model.layers.26.self_attn.v_experts.17", + "model.layers.26.self_attn.v_experts.18", + "model.layers.26.self_attn.v_experts.19", + "model.layers.26.self_attn.v_experts.20", + "model.layers.26.self_attn.v_experts.21", + "model.layers.26.self_attn.v_experts.22", + "model.layers.26.self_attn.v_experts.23", + "model.layers.26.self_attn.v_experts.24", + "model.layers.26.self_attn.v_experts.25", + "model.layers.26.self_attn.v_experts.26", + "model.layers.26.self_attn.v_experts.27", + "model.layers.26.self_attn.v_experts.28", + "model.layers.26.self_attn.v_experts.29", + "model.layers.26.self_attn.v_experts.30", + "model.layers.26.self_attn.v_experts.31", + "model.layers.26.self_attn.v_experts.32", + "model.layers.26.self_attn.v_experts.33", + "model.layers.26.self_attn.v_experts.34", + "model.layers.26.self_attn.v_experts.35", + "model.layers.26.self_attn.v_experts.36", + "model.layers.26.self_attn.v_experts.37", + "model.layers.26.self_attn.v_experts.38", + "model.layers.26.self_attn.v_experts.39", + "model.layers.26.self_attn.v_experts.40", + "model.layers.26.self_attn.v_experts.41", + "model.layers.26.self_attn.v_experts.42", + "model.layers.26.self_attn.v_experts.43", + "model.layers.26.self_attn.v_experts.44", + "model.layers.26.self_attn.v_experts.45", + "model.layers.26.self_attn.v_experts.46", + "model.layers.26.self_attn.v_experts.47", + "model.layers.26.self_attn.v_experts.48", + "model.layers.26.self_attn.v_experts.49", + "model.layers.26.self_attn.v_experts.50", + "model.layers.26.self_attn.v_experts.51", + "model.layers.26.self_attn.v_experts.52", + "model.layers.26.self_attn.v_experts.53", + "model.layers.26.self_attn.v_experts.54", + "model.layers.26.self_attn.v_experts.55", + "model.layers.26.self_attn.v_experts.56", + "model.layers.26.self_attn.v_experts.57", + "model.layers.26.self_attn.v_experts.58", + "model.layers.26.self_attn.v_experts.59", + "model.layers.26.self_attn.v_experts.60", + "model.layers.26.self_attn.v_experts.61", + "model.layers.26.self_attn.v_experts.62", + "model.layers.26.self_attn.v_experts.63", + "model.layers.26.self_attn.gate_proj", + "model.layers.26.mlp.gate", + "model.layers.26.mlp.shared_experts.gate_proj", + "model.layers.26.mlp.shared_experts.up_proj", + "model.layers.26.mlp.shared_experts.down_proj", + "model.layers.27.self_attn.q_proj", + "model.layers.27.self_attn.k_proj", + "model.layers.27.self_attn.o_proj", + "model.layers.27.self_attn.v_router", + "model.layers.27.self_attn.v_experts.0", + "model.layers.27.self_attn.v_experts.1", + "model.layers.27.self_attn.v_experts.2", + "model.layers.27.self_attn.v_experts.3", + "model.layers.27.self_attn.v_experts.4", + "model.layers.27.self_attn.v_experts.5", + "model.layers.27.self_attn.v_experts.6", + "model.layers.27.self_attn.v_experts.7", + "model.layers.27.self_attn.v_experts.8", + "model.layers.27.self_attn.v_experts.9", + "model.layers.27.self_attn.v_experts.10", + "model.layers.27.self_attn.v_experts.11", + "model.layers.27.self_attn.v_experts.12", + "model.layers.27.self_attn.v_experts.13", + "model.layers.27.self_attn.v_experts.14", + "model.layers.27.self_attn.v_experts.15", + "model.layers.27.self_attn.v_experts.16", + "model.layers.27.self_attn.v_experts.17", + "model.layers.27.self_attn.v_experts.18", + "model.layers.27.self_attn.v_experts.19", + "model.layers.27.self_attn.v_experts.20", + "model.layers.27.self_attn.v_experts.21", + "model.layers.27.self_attn.v_experts.22", + "model.layers.27.self_attn.v_experts.23", + "model.layers.27.self_attn.v_experts.24", + "model.layers.27.self_attn.v_experts.25", + "model.layers.27.self_attn.v_experts.26", + "model.layers.27.self_attn.v_experts.27", + "model.layers.27.self_attn.v_experts.28", + "model.layers.27.self_attn.v_experts.29", + "model.layers.27.self_attn.v_experts.30", + "model.layers.27.self_attn.v_experts.31", + "model.layers.27.self_attn.v_experts.32", + "model.layers.27.self_attn.v_experts.33", + "model.layers.27.self_attn.v_experts.34", + "model.layers.27.self_attn.v_experts.35", + "model.layers.27.self_attn.v_experts.36", + "model.layers.27.self_attn.v_experts.37", + "model.layers.27.self_attn.v_experts.38", + "model.layers.27.self_attn.v_experts.39", + "model.layers.27.self_attn.v_experts.40", + "model.layers.27.self_attn.v_experts.41", + "model.layers.27.self_attn.v_experts.42", + "model.layers.27.self_attn.v_experts.43", + "model.layers.27.self_attn.v_experts.44", + "model.layers.27.self_attn.v_experts.45", + "model.layers.27.self_attn.v_experts.46", + "model.layers.27.self_attn.v_experts.47", + "model.layers.27.self_attn.v_experts.48", + "model.layers.27.self_attn.v_experts.49", + "model.layers.27.self_attn.v_experts.50", + "model.layers.27.self_attn.v_experts.51", + "model.layers.27.self_attn.v_experts.52", + "model.layers.27.self_attn.v_experts.53", + "model.layers.27.self_attn.v_experts.54", + "model.layers.27.self_attn.v_experts.55", + "model.layers.27.self_attn.v_experts.56", + "model.layers.27.self_attn.v_experts.57", + "model.layers.27.self_attn.v_experts.58", + "model.layers.27.self_attn.v_experts.59", + "model.layers.27.self_attn.v_experts.60", + "model.layers.27.self_attn.v_experts.61", + "model.layers.27.self_attn.v_experts.62", + "model.layers.27.self_attn.v_experts.63", + "model.layers.27.self_attn.gate_proj", + "model.layers.27.mlp.gate", + "model.layers.27.mlp.shared_experts.gate_proj", + "model.layers.27.mlp.shared_experts.up_proj", + "model.layers.27.mlp.shared_experts.down_proj", + "model.layers.28.self_attn.q_proj", + "model.layers.28.self_attn.k_proj", + "model.layers.28.self_attn.o_proj", + "model.layers.28.self_attn.v_router", + "model.layers.28.self_attn.v_experts.0", + "model.layers.28.self_attn.v_experts.1", + "model.layers.28.self_attn.v_experts.2", + "model.layers.28.self_attn.v_experts.3", + "model.layers.28.self_attn.v_experts.4", + "model.layers.28.self_attn.v_experts.5", + "model.layers.28.self_attn.v_experts.6", + "model.layers.28.self_attn.v_experts.7", + "model.layers.28.self_attn.v_experts.8", + "model.layers.28.self_attn.v_experts.9", + "model.layers.28.self_attn.v_experts.10", + "model.layers.28.self_attn.v_experts.11", + "model.layers.28.self_attn.v_experts.12", + "model.layers.28.self_attn.v_experts.13", + "model.layers.28.self_attn.v_experts.14", + "model.layers.28.self_attn.v_experts.15", + "model.layers.28.self_attn.v_experts.16", + "model.layers.28.self_attn.v_experts.17", + "model.layers.28.self_attn.v_experts.18", + "model.layers.28.self_attn.v_experts.19", + "model.layers.28.self_attn.v_experts.20", + "model.layers.28.self_attn.v_experts.21", + "model.layers.28.self_attn.v_experts.22", + "model.layers.28.self_attn.v_experts.23", + "model.layers.28.self_attn.v_experts.24", + "model.layers.28.self_attn.v_experts.25", + "model.layers.28.self_attn.v_experts.26", + "model.layers.28.self_attn.v_experts.27", + "model.layers.28.self_attn.v_experts.28", + "model.layers.28.self_attn.v_experts.29", + "model.layers.28.self_attn.v_experts.30", + "model.layers.28.self_attn.v_experts.31", + "model.layers.28.self_attn.v_experts.32", + "model.layers.28.self_attn.v_experts.33", + "model.layers.28.self_attn.v_experts.34", + "model.layers.28.self_attn.v_experts.35", + "model.layers.28.self_attn.v_experts.36", + "model.layers.28.self_attn.v_experts.37", + "model.layers.28.self_attn.v_experts.38", + "model.layers.28.self_attn.v_experts.39", + "model.layers.28.self_attn.v_experts.40", + "model.layers.28.self_attn.v_experts.41", + "model.layers.28.self_attn.v_experts.42", + "model.layers.28.self_attn.v_experts.43", + "model.layers.28.self_attn.v_experts.44", + "model.layers.28.self_attn.v_experts.45", + "model.layers.28.self_attn.v_experts.46", + "model.layers.28.self_attn.v_experts.47", + "model.layers.28.self_attn.v_experts.48", + "model.layers.28.self_attn.v_experts.49", + "model.layers.28.self_attn.v_experts.50", + "model.layers.28.self_attn.v_experts.51", + "model.layers.28.self_attn.v_experts.52", + "model.layers.28.self_attn.v_experts.53", + "model.layers.28.self_attn.v_experts.54", + "model.layers.28.self_attn.v_experts.55", + "model.layers.28.self_attn.v_experts.56", + "model.layers.28.self_attn.v_experts.57", + "model.layers.28.self_attn.v_experts.58", + "model.layers.28.self_attn.v_experts.59", + "model.layers.28.self_attn.v_experts.60", + "model.layers.28.self_attn.v_experts.61", + "model.layers.28.self_attn.v_experts.62", + "model.layers.28.self_attn.v_experts.63", + "model.layers.28.self_attn.gate_proj", + "model.layers.28.mlp.gate", + "model.layers.28.mlp.shared_experts.gate_proj", + "model.layers.28.mlp.shared_experts.up_proj", + "model.layers.28.mlp.shared_experts.down_proj", + "model.layers.29.self_attn.q_proj", + "model.layers.29.self_attn.k_proj", + "model.layers.29.self_attn.o_proj", + "model.layers.29.self_attn.v_router", + "model.layers.29.self_attn.v_experts.0", + "model.layers.29.self_attn.v_experts.1", + "model.layers.29.self_attn.v_experts.2", + "model.layers.29.self_attn.v_experts.3", + "model.layers.29.self_attn.v_experts.4", + "model.layers.29.self_attn.v_experts.5", + "model.layers.29.self_attn.v_experts.6", + "model.layers.29.self_attn.v_experts.7", + "model.layers.29.self_attn.v_experts.8", + "model.layers.29.self_attn.v_experts.9", + "model.layers.29.self_attn.v_experts.10", + "model.layers.29.self_attn.v_experts.11", + "model.layers.29.self_attn.v_experts.12", + "model.layers.29.self_attn.v_experts.13", + "model.layers.29.self_attn.v_experts.14", + "model.layers.29.self_attn.v_experts.15", + "model.layers.29.self_attn.v_experts.16", + "model.layers.29.self_attn.v_experts.17", + "model.layers.29.self_attn.v_experts.18", + "model.layers.29.self_attn.v_experts.19", + "model.layers.29.self_attn.v_experts.20", + "model.layers.29.self_attn.v_experts.21", + "model.layers.29.self_attn.v_experts.22", + "model.layers.29.self_attn.v_experts.23", + "model.layers.29.self_attn.v_experts.24", + "model.layers.29.self_attn.v_experts.25", + "model.layers.29.self_attn.v_experts.26", + "model.layers.29.self_attn.v_experts.27", + "model.layers.29.self_attn.v_experts.28", + "model.layers.29.self_attn.v_experts.29", + "model.layers.29.self_attn.v_experts.30", + "model.layers.29.self_attn.v_experts.31", + "model.layers.29.self_attn.v_experts.32", + "model.layers.29.self_attn.v_experts.33", + "model.layers.29.self_attn.v_experts.34", + "model.layers.29.self_attn.v_experts.35", + "model.layers.29.self_attn.v_experts.36", + "model.layers.29.self_attn.v_experts.37", + "model.layers.29.self_attn.v_experts.38", + "model.layers.29.self_attn.v_experts.39", + "model.layers.29.self_attn.v_experts.40", + "model.layers.29.self_attn.v_experts.41", + "model.layers.29.self_attn.v_experts.42", + "model.layers.29.self_attn.v_experts.43", + "model.layers.29.self_attn.v_experts.44", + "model.layers.29.self_attn.v_experts.45", + "model.layers.29.self_attn.v_experts.46", + "model.layers.29.self_attn.v_experts.47", + "model.layers.29.self_attn.v_experts.48", + "model.layers.29.self_attn.v_experts.49", + "model.layers.29.self_attn.v_experts.50", + "model.layers.29.self_attn.v_experts.51", + "model.layers.29.self_attn.v_experts.52", + "model.layers.29.self_attn.v_experts.53", + "model.layers.29.self_attn.v_experts.54", + "model.layers.29.self_attn.v_experts.55", + "model.layers.29.self_attn.v_experts.56", + "model.layers.29.self_attn.v_experts.57", + "model.layers.29.self_attn.v_experts.58", + "model.layers.29.self_attn.v_experts.59", + "model.layers.29.self_attn.v_experts.60", + "model.layers.29.self_attn.v_experts.61", + "model.layers.29.self_attn.v_experts.62", + "model.layers.29.self_attn.v_experts.63", + "model.layers.29.self_attn.gate_proj", + "model.layers.29.mlp.gate", + "model.layers.29.mlp.shared_experts.gate_proj", + "model.layers.29.mlp.shared_experts.up_proj", + "model.layers.29.mlp.shared_experts.down_proj", + "model.layers.30.self_attn.q_proj", + "model.layers.30.self_attn.k_proj", + "model.layers.30.self_attn.o_proj", + "model.layers.30.self_attn.v_router", + "model.layers.30.self_attn.v_experts.0", + "model.layers.30.self_attn.v_experts.1", + "model.layers.30.self_attn.v_experts.2", + "model.layers.30.self_attn.v_experts.3", + "model.layers.30.self_attn.v_experts.4", + "model.layers.30.self_attn.v_experts.5", + "model.layers.30.self_attn.v_experts.6", + "model.layers.30.self_attn.v_experts.7", + "model.layers.30.self_attn.v_experts.8", + "model.layers.30.self_attn.v_experts.9", + "model.layers.30.self_attn.v_experts.10", + "model.layers.30.self_attn.v_experts.11", + "model.layers.30.self_attn.v_experts.12", + "model.layers.30.self_attn.v_experts.13", + "model.layers.30.self_attn.v_experts.14", + "model.layers.30.self_attn.v_experts.15", + "model.layers.30.self_attn.v_experts.16", + "model.layers.30.self_attn.v_experts.17", + "model.layers.30.self_attn.v_experts.18", + "model.layers.30.self_attn.v_experts.19", + "model.layers.30.self_attn.v_experts.20", + "model.layers.30.self_attn.v_experts.21", + "model.layers.30.self_attn.v_experts.22", + "model.layers.30.self_attn.v_experts.23", + "model.layers.30.self_attn.v_experts.24", + "model.layers.30.self_attn.v_experts.25", + "model.layers.30.self_attn.v_experts.26", + "model.layers.30.self_attn.v_experts.27", + "model.layers.30.self_attn.v_experts.28", + "model.layers.30.self_attn.v_experts.29", + "model.layers.30.self_attn.v_experts.30", + "model.layers.30.self_attn.v_experts.31", + "model.layers.30.self_attn.v_experts.32", + "model.layers.30.self_attn.v_experts.33", + "model.layers.30.self_attn.v_experts.34", + "model.layers.30.self_attn.v_experts.35", + "model.layers.30.self_attn.v_experts.36", + "model.layers.30.self_attn.v_experts.37", + "model.layers.30.self_attn.v_experts.38", + "model.layers.30.self_attn.v_experts.39", + "model.layers.30.self_attn.v_experts.40", + "model.layers.30.self_attn.v_experts.41", + "model.layers.30.self_attn.v_experts.42", + "model.layers.30.self_attn.v_experts.43", + "model.layers.30.self_attn.v_experts.44", + "model.layers.30.self_attn.v_experts.45", + "model.layers.30.self_attn.v_experts.46", + "model.layers.30.self_attn.v_experts.47", + "model.layers.30.self_attn.v_experts.48", + "model.layers.30.self_attn.v_experts.49", + "model.layers.30.self_attn.v_experts.50", + "model.layers.30.self_attn.v_experts.51", + "model.layers.30.self_attn.v_experts.52", + "model.layers.30.self_attn.v_experts.53", + "model.layers.30.self_attn.v_experts.54", + "model.layers.30.self_attn.v_experts.55", + "model.layers.30.self_attn.v_experts.56", + "model.layers.30.self_attn.v_experts.57", + "model.layers.30.self_attn.v_experts.58", + "model.layers.30.self_attn.v_experts.59", + "model.layers.30.self_attn.v_experts.60", + "model.layers.30.self_attn.v_experts.61", + "model.layers.30.self_attn.v_experts.62", + "model.layers.30.self_attn.v_experts.63", + "model.layers.30.self_attn.gate_proj", + "model.layers.30.mlp.gate", + "model.layers.30.mlp.shared_experts.gate_proj", + "model.layers.30.mlp.shared_experts.up_proj", + "model.layers.30.mlp.shared_experts.down_proj", + "model.layers.31.self_attn.q_proj", + "model.layers.31.self_attn.k_proj", + "model.layers.31.self_attn.o_proj", + "model.layers.31.self_attn.v_router", + "model.layers.31.self_attn.v_experts.0", + "model.layers.31.self_attn.v_experts.1", + "model.layers.31.self_attn.v_experts.2", + "model.layers.31.self_attn.v_experts.3", + "model.layers.31.self_attn.v_experts.4", + "model.layers.31.self_attn.v_experts.5", + "model.layers.31.self_attn.v_experts.6", + "model.layers.31.self_attn.v_experts.7", + "model.layers.31.self_attn.v_experts.8", + "model.layers.31.self_attn.v_experts.9", + "model.layers.31.self_attn.v_experts.10", + "model.layers.31.self_attn.v_experts.11", + "model.layers.31.self_attn.v_experts.12", + "model.layers.31.self_attn.v_experts.13", + "model.layers.31.self_attn.v_experts.14", + "model.layers.31.self_attn.v_experts.15", + "model.layers.31.self_attn.v_experts.16", + "model.layers.31.self_attn.v_experts.17", + "model.layers.31.self_attn.v_experts.18", + "model.layers.31.self_attn.v_experts.19", + "model.layers.31.self_attn.v_experts.20", + "model.layers.31.self_attn.v_experts.21", + "model.layers.31.self_attn.v_experts.22", + "model.layers.31.self_attn.v_experts.23", + "model.layers.31.self_attn.v_experts.24", + "model.layers.31.self_attn.v_experts.25", + "model.layers.31.self_attn.v_experts.26", + "model.layers.31.self_attn.v_experts.27", + "model.layers.31.self_attn.v_experts.28", + "model.layers.31.self_attn.v_experts.29", + "model.layers.31.self_attn.v_experts.30", + "model.layers.31.self_attn.v_experts.31", + "model.layers.31.self_attn.v_experts.32", + "model.layers.31.self_attn.v_experts.33", + "model.layers.31.self_attn.v_experts.34", + "model.layers.31.self_attn.v_experts.35", + "model.layers.31.self_attn.v_experts.36", + "model.layers.31.self_attn.v_experts.37", + "model.layers.31.self_attn.v_experts.38", + "model.layers.31.self_attn.v_experts.39", + "model.layers.31.self_attn.v_experts.40", + "model.layers.31.self_attn.v_experts.41", + "model.layers.31.self_attn.v_experts.42", + "model.layers.31.self_attn.v_experts.43", + "model.layers.31.self_attn.v_experts.44", + "model.layers.31.self_attn.v_experts.45", + "model.layers.31.self_attn.v_experts.46", + "model.layers.31.self_attn.v_experts.47", + "model.layers.31.self_attn.v_experts.48", + "model.layers.31.self_attn.v_experts.49", + "model.layers.31.self_attn.v_experts.50", + "model.layers.31.self_attn.v_experts.51", + "model.layers.31.self_attn.v_experts.52", + "model.layers.31.self_attn.v_experts.53", + "model.layers.31.self_attn.v_experts.54", + "model.layers.31.self_attn.v_experts.55", + "model.layers.31.self_attn.v_experts.56", + "model.layers.31.self_attn.v_experts.57", + "model.layers.31.self_attn.v_experts.58", + "model.layers.31.self_attn.v_experts.59", + "model.layers.31.self_attn.v_experts.60", + "model.layers.31.self_attn.v_experts.61", + "model.layers.31.self_attn.v_experts.62", + "model.layers.31.self_attn.v_experts.63", + "model.layers.31.self_attn.gate_proj", + "model.layers.31.mlp.gate", + "model.layers.31.mlp.shared_experts.gate_proj", + "model.layers.31.mlp.shared_experts.up_proj", + "model.layers.31.mlp.shared_experts.down_proj", + "model.layers.32.self_attn.q_proj", + "model.layers.32.self_attn.k_proj", + "model.layers.32.self_attn.o_proj", + "model.layers.32.self_attn.v_router", + "model.layers.32.self_attn.v_experts.0", + "model.layers.32.self_attn.v_experts.1", + "model.layers.32.self_attn.v_experts.2", + "model.layers.32.self_attn.v_experts.3", + "model.layers.32.self_attn.v_experts.4", + "model.layers.32.self_attn.v_experts.5", + "model.layers.32.self_attn.v_experts.6", + "model.layers.32.self_attn.v_experts.7", + "model.layers.32.self_attn.v_experts.8", + "model.layers.32.self_attn.v_experts.9", + "model.layers.32.self_attn.v_experts.10", + "model.layers.32.self_attn.v_experts.11", + "model.layers.32.self_attn.v_experts.12", + "model.layers.32.self_attn.v_experts.13", + "model.layers.32.self_attn.v_experts.14", + "model.layers.32.self_attn.v_experts.15", + "model.layers.32.self_attn.v_experts.16", + "model.layers.32.self_attn.v_experts.17", + "model.layers.32.self_attn.v_experts.18", + "model.layers.32.self_attn.v_experts.19", + "model.layers.32.self_attn.v_experts.20", + "model.layers.32.self_attn.v_experts.21", + "model.layers.32.self_attn.v_experts.22", + "model.layers.32.self_attn.v_experts.23", + "model.layers.32.self_attn.v_experts.24", + "model.layers.32.self_attn.v_experts.25", + "model.layers.32.self_attn.v_experts.26", + "model.layers.32.self_attn.v_experts.27", + "model.layers.32.self_attn.v_experts.28", + "model.layers.32.self_attn.v_experts.29", + "model.layers.32.self_attn.v_experts.30", + "model.layers.32.self_attn.v_experts.31", + "model.layers.32.self_attn.v_experts.32", + "model.layers.32.self_attn.v_experts.33", + "model.layers.32.self_attn.v_experts.34", + "model.layers.32.self_attn.v_experts.35", + "model.layers.32.self_attn.v_experts.36", + "model.layers.32.self_attn.v_experts.37", + "model.layers.32.self_attn.v_experts.38", + "model.layers.32.self_attn.v_experts.39", + "model.layers.32.self_attn.v_experts.40", + "model.layers.32.self_attn.v_experts.41", + "model.layers.32.self_attn.v_experts.42", + "model.layers.32.self_attn.v_experts.43", + "model.layers.32.self_attn.v_experts.44", + "model.layers.32.self_attn.v_experts.45", + "model.layers.32.self_attn.v_experts.46", + "model.layers.32.self_attn.v_experts.47", + "model.layers.32.self_attn.v_experts.48", + "model.layers.32.self_attn.v_experts.49", + "model.layers.32.self_attn.v_experts.50", + "model.layers.32.self_attn.v_experts.51", + "model.layers.32.self_attn.v_experts.52", + "model.layers.32.self_attn.v_experts.53", + "model.layers.32.self_attn.v_experts.54", + "model.layers.32.self_attn.v_experts.55", + "model.layers.32.self_attn.v_experts.56", + "model.layers.32.self_attn.v_experts.57", + "model.layers.32.self_attn.v_experts.58", + "model.layers.32.self_attn.v_experts.59", + "model.layers.32.self_attn.v_experts.60", + "model.layers.32.self_attn.v_experts.61", + "model.layers.32.self_attn.v_experts.62", + "model.layers.32.self_attn.v_experts.63", + "model.layers.32.self_attn.gate_proj", + "model.layers.32.mlp.gate", + "model.layers.32.mlp.shared_experts.gate_proj", + "model.layers.32.mlp.shared_experts.up_proj", + "model.layers.32.mlp.shared_experts.down_proj", + "model.layers.33.self_attn.q_proj", + "model.layers.33.self_attn.k_proj", + "model.layers.33.self_attn.o_proj", + "model.layers.33.self_attn.v_router", + "model.layers.33.self_attn.v_experts.0", + "model.layers.33.self_attn.v_experts.1", + "model.layers.33.self_attn.v_experts.2", + "model.layers.33.self_attn.v_experts.3", + "model.layers.33.self_attn.v_experts.4", + "model.layers.33.self_attn.v_experts.5", + "model.layers.33.self_attn.v_experts.6", + "model.layers.33.self_attn.v_experts.7", + "model.layers.33.self_attn.v_experts.8", + "model.layers.33.self_attn.v_experts.9", + "model.layers.33.self_attn.v_experts.10", + "model.layers.33.self_attn.v_experts.11", + "model.layers.33.self_attn.v_experts.12", + "model.layers.33.self_attn.v_experts.13", + "model.layers.33.self_attn.v_experts.14", + "model.layers.33.self_attn.v_experts.15", + "model.layers.33.self_attn.v_experts.16", + "model.layers.33.self_attn.v_experts.17", + "model.layers.33.self_attn.v_experts.18", + "model.layers.33.self_attn.v_experts.19", + "model.layers.33.self_attn.v_experts.20", + "model.layers.33.self_attn.v_experts.21", + "model.layers.33.self_attn.v_experts.22", + "model.layers.33.self_attn.v_experts.23", + "model.layers.33.self_attn.v_experts.24", + "model.layers.33.self_attn.v_experts.25", + "model.layers.33.self_attn.v_experts.26", + "model.layers.33.self_attn.v_experts.27", + "model.layers.33.self_attn.v_experts.28", + "model.layers.33.self_attn.v_experts.29", + "model.layers.33.self_attn.v_experts.30", + "model.layers.33.self_attn.v_experts.31", + "model.layers.33.self_attn.v_experts.32", + "model.layers.33.self_attn.v_experts.33", + "model.layers.33.self_attn.v_experts.34", + "model.layers.33.self_attn.v_experts.35", + "model.layers.33.self_attn.v_experts.36", + "model.layers.33.self_attn.v_experts.37", + "model.layers.33.self_attn.v_experts.38", + "model.layers.33.self_attn.v_experts.39", + "model.layers.33.self_attn.v_experts.40", + "model.layers.33.self_attn.v_experts.41", + "model.layers.33.self_attn.v_experts.42", + "model.layers.33.self_attn.v_experts.43", + "model.layers.33.self_attn.v_experts.44", + "model.layers.33.self_attn.v_experts.45", + "model.layers.33.self_attn.v_experts.46", + "model.layers.33.self_attn.v_experts.47", + "model.layers.33.self_attn.v_experts.48", + "model.layers.33.self_attn.v_experts.49", + "model.layers.33.self_attn.v_experts.50", + "model.layers.33.self_attn.v_experts.51", + "model.layers.33.self_attn.v_experts.52", + "model.layers.33.self_attn.v_experts.53", + "model.layers.33.self_attn.v_experts.54", + "model.layers.33.self_attn.v_experts.55", + "model.layers.33.self_attn.v_experts.56", + "model.layers.33.self_attn.v_experts.57", + "model.layers.33.self_attn.v_experts.58", + "model.layers.33.self_attn.v_experts.59", + "model.layers.33.self_attn.v_experts.60", + "model.layers.33.self_attn.v_experts.61", + "model.layers.33.self_attn.v_experts.62", + "model.layers.33.self_attn.v_experts.63", + "model.layers.33.self_attn.gate_proj", + "model.layers.33.mlp.gate", + "model.layers.33.mlp.shared_experts.gate_proj", + "model.layers.33.mlp.shared_experts.up_proj", + "model.layers.33.mlp.shared_experts.down_proj", + "model.layers.34.self_attn.q_proj", + "model.layers.34.self_attn.k_proj", + "model.layers.34.self_attn.o_proj", + "model.layers.34.self_attn.v_router", + "model.layers.34.self_attn.v_experts.0", + "model.layers.34.self_attn.v_experts.1", + "model.layers.34.self_attn.v_experts.2", + "model.layers.34.self_attn.v_experts.3", + "model.layers.34.self_attn.v_experts.4", + "model.layers.34.self_attn.v_experts.5", + "model.layers.34.self_attn.v_experts.6", + "model.layers.34.self_attn.v_experts.7", + "model.layers.34.self_attn.v_experts.8", + "model.layers.34.self_attn.v_experts.9", + "model.layers.34.self_attn.v_experts.10", + "model.layers.34.self_attn.v_experts.11", + "model.layers.34.self_attn.v_experts.12", + "model.layers.34.self_attn.v_experts.13", + "model.layers.34.self_attn.v_experts.14", + "model.layers.34.self_attn.v_experts.15", + "model.layers.34.self_attn.v_experts.16", + "model.layers.34.self_attn.v_experts.17", + "model.layers.34.self_attn.v_experts.18", + "model.layers.34.self_attn.v_experts.19", + "model.layers.34.self_attn.v_experts.20", + "model.layers.34.self_attn.v_experts.21", + "model.layers.34.self_attn.v_experts.22", + "model.layers.34.self_attn.v_experts.23", + "model.layers.34.self_attn.v_experts.24", + "model.layers.34.self_attn.v_experts.25", + "model.layers.34.self_attn.v_experts.26", + "model.layers.34.self_attn.v_experts.27", + "model.layers.34.self_attn.v_experts.28", + "model.layers.34.self_attn.v_experts.29", + "model.layers.34.self_attn.v_experts.30", + "model.layers.34.self_attn.v_experts.31", + "model.layers.34.self_attn.v_experts.32", + "model.layers.34.self_attn.v_experts.33", + "model.layers.34.self_attn.v_experts.34", + "model.layers.34.self_attn.v_experts.35", + "model.layers.34.self_attn.v_experts.36", + "model.layers.34.self_attn.v_experts.37", + "model.layers.34.self_attn.v_experts.38", + "model.layers.34.self_attn.v_experts.39", + "model.layers.34.self_attn.v_experts.40", + "model.layers.34.self_attn.v_experts.41", + "model.layers.34.self_attn.v_experts.42", + "model.layers.34.self_attn.v_experts.43", + "model.layers.34.self_attn.v_experts.44", + "model.layers.34.self_attn.v_experts.45", + "model.layers.34.self_attn.v_experts.46", + "model.layers.34.self_attn.v_experts.47", + "model.layers.34.self_attn.v_experts.48", + "model.layers.34.self_attn.v_experts.49", + "model.layers.34.self_attn.v_experts.50", + "model.layers.34.self_attn.v_experts.51", + "model.layers.34.self_attn.v_experts.52", + "model.layers.34.self_attn.v_experts.53", + "model.layers.34.self_attn.v_experts.54", + "model.layers.34.self_attn.v_experts.55", + "model.layers.34.self_attn.v_experts.56", + "model.layers.34.self_attn.v_experts.57", + "model.layers.34.self_attn.v_experts.58", + "model.layers.34.self_attn.v_experts.59", + "model.layers.34.self_attn.v_experts.60", + "model.layers.34.self_attn.v_experts.61", + "model.layers.34.self_attn.v_experts.62", + "model.layers.34.self_attn.v_experts.63", + "model.layers.34.self_attn.gate_proj", + "model.layers.34.mlp.gate", + "model.layers.34.mlp.shared_experts.gate_proj", + "model.layers.34.mlp.shared_experts.up_proj", + "model.layers.34.mlp.shared_experts.down_proj", + "model.layers.35.self_attn.q_proj", + "model.layers.35.self_attn.k_proj", + "model.layers.35.self_attn.o_proj", + "model.layers.35.self_attn.v_router", + "model.layers.35.self_attn.v_experts.0", + "model.layers.35.self_attn.v_experts.1", + "model.layers.35.self_attn.v_experts.2", + "model.layers.35.self_attn.v_experts.3", + "model.layers.35.self_attn.v_experts.4", + "model.layers.35.self_attn.v_experts.5", + "model.layers.35.self_attn.v_experts.6", + "model.layers.35.self_attn.v_experts.7", + "model.layers.35.self_attn.v_experts.8", + "model.layers.35.self_attn.v_experts.9", + "model.layers.35.self_attn.v_experts.10", + "model.layers.35.self_attn.v_experts.11", + "model.layers.35.self_attn.v_experts.12", + "model.layers.35.self_attn.v_experts.13", + "model.layers.35.self_attn.v_experts.14", + "model.layers.35.self_attn.v_experts.15", + "model.layers.35.self_attn.v_experts.16", + "model.layers.35.self_attn.v_experts.17", + "model.layers.35.self_attn.v_experts.18", + "model.layers.35.self_attn.v_experts.19", + "model.layers.35.self_attn.v_experts.20", + "model.layers.35.self_attn.v_experts.21", + "model.layers.35.self_attn.v_experts.22", + "model.layers.35.self_attn.v_experts.23", + "model.layers.35.self_attn.v_experts.24", + "model.layers.35.self_attn.v_experts.25", + "model.layers.35.self_attn.v_experts.26", + "model.layers.35.self_attn.v_experts.27", + "model.layers.35.self_attn.v_experts.28", + "model.layers.35.self_attn.v_experts.29", + "model.layers.35.self_attn.v_experts.30", + "model.layers.35.self_attn.v_experts.31", + "model.layers.35.self_attn.v_experts.32", + "model.layers.35.self_attn.v_experts.33", + "model.layers.35.self_attn.v_experts.34", + "model.layers.35.self_attn.v_experts.35", + "model.layers.35.self_attn.v_experts.36", + "model.layers.35.self_attn.v_experts.37", + "model.layers.35.self_attn.v_experts.38", + "model.layers.35.self_attn.v_experts.39", + "model.layers.35.self_attn.v_experts.40", + "model.layers.35.self_attn.v_experts.41", + "model.layers.35.self_attn.v_experts.42", + "model.layers.35.self_attn.v_experts.43", + "model.layers.35.self_attn.v_experts.44", + "model.layers.35.self_attn.v_experts.45", + "model.layers.35.self_attn.v_experts.46", + "model.layers.35.self_attn.v_experts.47", + "model.layers.35.self_attn.v_experts.48", + "model.layers.35.self_attn.v_experts.49", + "model.layers.35.self_attn.v_experts.50", + "model.layers.35.self_attn.v_experts.51", + "model.layers.35.self_attn.v_experts.52", + "model.layers.35.self_attn.v_experts.53", + "model.layers.35.self_attn.v_experts.54", + "model.layers.35.self_attn.v_experts.55", + "model.layers.35.self_attn.v_experts.56", + "model.layers.35.self_attn.v_experts.57", + "model.layers.35.self_attn.v_experts.58", + "model.layers.35.self_attn.v_experts.59", + "model.layers.35.self_attn.v_experts.60", + "model.layers.35.self_attn.v_experts.61", + "model.layers.35.self_attn.v_experts.62", + "model.layers.35.self_attn.v_experts.63", + "model.layers.35.self_attn.gate_proj", + "model.layers.35.mlp.gate", + "model.layers.35.mlp.shared_experts.gate_proj", + "model.layers.35.mlp.shared_experts.up_proj", + "model.layers.35.mlp.shared_experts.down_proj", + "model.layers.36.self_attn.q_proj", + "model.layers.36.self_attn.k_proj", + "model.layers.36.self_attn.o_proj", + "model.layers.36.self_attn.v_router", + "model.layers.36.self_attn.v_experts.0", + "model.layers.36.self_attn.v_experts.1", + "model.layers.36.self_attn.v_experts.2", + "model.layers.36.self_attn.v_experts.3", + "model.layers.36.self_attn.v_experts.4", + "model.layers.36.self_attn.v_experts.5", + "model.layers.36.self_attn.v_experts.6", + "model.layers.36.self_attn.v_experts.7", + "model.layers.36.self_attn.v_experts.8", + "model.layers.36.self_attn.v_experts.9", + "model.layers.36.self_attn.v_experts.10", + "model.layers.36.self_attn.v_experts.11", + "model.layers.36.self_attn.v_experts.12", + "model.layers.36.self_attn.v_experts.13", + "model.layers.36.self_attn.v_experts.14", + "model.layers.36.self_attn.v_experts.15", + "model.layers.36.self_attn.v_experts.16", + "model.layers.36.self_attn.v_experts.17", + "model.layers.36.self_attn.v_experts.18", + "model.layers.36.self_attn.v_experts.19", + "model.layers.36.self_attn.v_experts.20", + "model.layers.36.self_attn.v_experts.21", + "model.layers.36.self_attn.v_experts.22", + "model.layers.36.self_attn.v_experts.23", + "model.layers.36.self_attn.v_experts.24", + "model.layers.36.self_attn.v_experts.25", + "model.layers.36.self_attn.v_experts.26", + "model.layers.36.self_attn.v_experts.27", + "model.layers.36.self_attn.v_experts.28", + "model.layers.36.self_attn.v_experts.29", + "model.layers.36.self_attn.v_experts.30", + "model.layers.36.self_attn.v_experts.31", + "model.layers.36.self_attn.v_experts.32", + "model.layers.36.self_attn.v_experts.33", + "model.layers.36.self_attn.v_experts.34", + "model.layers.36.self_attn.v_experts.35", + "model.layers.36.self_attn.v_experts.36", + "model.layers.36.self_attn.v_experts.37", + "model.layers.36.self_attn.v_experts.38", + "model.layers.36.self_attn.v_experts.39", + "model.layers.36.self_attn.v_experts.40", + "model.layers.36.self_attn.v_experts.41", + "model.layers.36.self_attn.v_experts.42", + "model.layers.36.self_attn.v_experts.43", + "model.layers.36.self_attn.v_experts.44", + "model.layers.36.self_attn.v_experts.45", + "model.layers.36.self_attn.v_experts.46", + "model.layers.36.self_attn.v_experts.47", + "model.layers.36.self_attn.v_experts.48", + "model.layers.36.self_attn.v_experts.49", + "model.layers.36.self_attn.v_experts.50", + "model.layers.36.self_attn.v_experts.51", + "model.layers.36.self_attn.v_experts.52", + "model.layers.36.self_attn.v_experts.53", + "model.layers.36.self_attn.v_experts.54", + "model.layers.36.self_attn.v_experts.55", + "model.layers.36.self_attn.v_experts.56", + "model.layers.36.self_attn.v_experts.57", + "model.layers.36.self_attn.v_experts.58", + "model.layers.36.self_attn.v_experts.59", + "model.layers.36.self_attn.v_experts.60", + "model.layers.36.self_attn.v_experts.61", + "model.layers.36.self_attn.v_experts.62", + "model.layers.36.self_attn.v_experts.63", + "model.layers.36.self_attn.gate_proj", + "model.layers.36.mlp.gate", + "model.layers.36.mlp.shared_experts.gate_proj", + "model.layers.36.mlp.shared_experts.up_proj", + "model.layers.36.mlp.shared_experts.down_proj", + "model.layers.37.self_attn.q_proj", + "model.layers.37.self_attn.k_proj", + "model.layers.37.self_attn.o_proj", + "model.layers.37.self_attn.v_router", + "model.layers.37.self_attn.v_experts.0", + "model.layers.37.self_attn.v_experts.1", + "model.layers.37.self_attn.v_experts.2", + "model.layers.37.self_attn.v_experts.3", + "model.layers.37.self_attn.v_experts.4", + "model.layers.37.self_attn.v_experts.5", + "model.layers.37.self_attn.v_experts.6", + "model.layers.37.self_attn.v_experts.7", + "model.layers.37.self_attn.v_experts.8", + "model.layers.37.self_attn.v_experts.9", + "model.layers.37.self_attn.v_experts.10", + "model.layers.37.self_attn.v_experts.11", + "model.layers.37.self_attn.v_experts.12", + "model.layers.37.self_attn.v_experts.13", + "model.layers.37.self_attn.v_experts.14", + "model.layers.37.self_attn.v_experts.15", + "model.layers.37.self_attn.v_experts.16", + "model.layers.37.self_attn.v_experts.17", + "model.layers.37.self_attn.v_experts.18", + "model.layers.37.self_attn.v_experts.19", + "model.layers.37.self_attn.v_experts.20", + "model.layers.37.self_attn.v_experts.21", + "model.layers.37.self_attn.v_experts.22", + "model.layers.37.self_attn.v_experts.23", + "model.layers.37.self_attn.v_experts.24", + "model.layers.37.self_attn.v_experts.25", + "model.layers.37.self_attn.v_experts.26", + "model.layers.37.self_attn.v_experts.27", + "model.layers.37.self_attn.v_experts.28", + "model.layers.37.self_attn.v_experts.29", + "model.layers.37.self_attn.v_experts.30", + "model.layers.37.self_attn.v_experts.31", + "model.layers.37.self_attn.v_experts.32", + "model.layers.37.self_attn.v_experts.33", + "model.layers.37.self_attn.v_experts.34", + "model.layers.37.self_attn.v_experts.35", + "model.layers.37.self_attn.v_experts.36", + "model.layers.37.self_attn.v_experts.37", + "model.layers.37.self_attn.v_experts.38", + "model.layers.37.self_attn.v_experts.39", + "model.layers.37.self_attn.v_experts.40", + "model.layers.37.self_attn.v_experts.41", + "model.layers.37.self_attn.v_experts.42", + "model.layers.37.self_attn.v_experts.43", + "model.layers.37.self_attn.v_experts.44", + "model.layers.37.self_attn.v_experts.45", + "model.layers.37.self_attn.v_experts.46", + "model.layers.37.self_attn.v_experts.47", + "model.layers.37.self_attn.v_experts.48", + "model.layers.37.self_attn.v_experts.49", + "model.layers.37.self_attn.v_experts.50", + "model.layers.37.self_attn.v_experts.51", + "model.layers.37.self_attn.v_experts.52", + "model.layers.37.self_attn.v_experts.53", + "model.layers.37.self_attn.v_experts.54", + "model.layers.37.self_attn.v_experts.55", + "model.layers.37.self_attn.v_experts.56", + "model.layers.37.self_attn.v_experts.57", + "model.layers.37.self_attn.v_experts.58", + "model.layers.37.self_attn.v_experts.59", + "model.layers.37.self_attn.v_experts.60", + "model.layers.37.self_attn.v_experts.61", + "model.layers.37.self_attn.v_experts.62", + "model.layers.37.self_attn.v_experts.63", + "model.layers.37.self_attn.gate_proj", + "model.layers.37.mlp.gate", + "model.layers.37.mlp.shared_experts.gate_proj", + "model.layers.37.mlp.shared_experts.up_proj", + "model.layers.37.mlp.shared_experts.down_proj", + "model.layers.38.self_attn.q_proj", + "model.layers.38.self_attn.k_proj", + "model.layers.38.self_attn.o_proj", + "model.layers.38.self_attn.v_router", + "model.layers.38.self_attn.v_experts.0", + "model.layers.38.self_attn.v_experts.1", + "model.layers.38.self_attn.v_experts.2", + "model.layers.38.self_attn.v_experts.3", + "model.layers.38.self_attn.v_experts.4", + "model.layers.38.self_attn.v_experts.5", + "model.layers.38.self_attn.v_experts.6", + "model.layers.38.self_attn.v_experts.7", + "model.layers.38.self_attn.v_experts.8", + "model.layers.38.self_attn.v_experts.9", + "model.layers.38.self_attn.v_experts.10", + "model.layers.38.self_attn.v_experts.11", + "model.layers.38.self_attn.v_experts.12", + "model.layers.38.self_attn.v_experts.13", + "model.layers.38.self_attn.v_experts.14", + "model.layers.38.self_attn.v_experts.15", + "model.layers.38.self_attn.v_experts.16", + "model.layers.38.self_attn.v_experts.17", + "model.layers.38.self_attn.v_experts.18", + "model.layers.38.self_attn.v_experts.19", + "model.layers.38.self_attn.v_experts.20", + "model.layers.38.self_attn.v_experts.21", + "model.layers.38.self_attn.v_experts.22", + "model.layers.38.self_attn.v_experts.23", + "model.layers.38.self_attn.v_experts.24", + "model.layers.38.self_attn.v_experts.25", + "model.layers.38.self_attn.v_experts.26", + "model.layers.38.self_attn.v_experts.27", + "model.layers.38.self_attn.v_experts.28", + "model.layers.38.self_attn.v_experts.29", + "model.layers.38.self_attn.v_experts.30", + "model.layers.38.self_attn.v_experts.31", + "model.layers.38.self_attn.v_experts.32", + "model.layers.38.self_attn.v_experts.33", + "model.layers.38.self_attn.v_experts.34", + "model.layers.38.self_attn.v_experts.35", + "model.layers.38.self_attn.v_experts.36", + "model.layers.38.self_attn.v_experts.37", + "model.layers.38.self_attn.v_experts.38", + "model.layers.38.self_attn.v_experts.39", + "model.layers.38.self_attn.v_experts.40", + "model.layers.38.self_attn.v_experts.41", + "model.layers.38.self_attn.v_experts.42", + "model.layers.38.self_attn.v_experts.43", + "model.layers.38.self_attn.v_experts.44", + "model.layers.38.self_attn.v_experts.45", + "model.layers.38.self_attn.v_experts.46", + "model.layers.38.self_attn.v_experts.47", + "model.layers.38.self_attn.v_experts.48", + "model.layers.38.self_attn.v_experts.49", + "model.layers.38.self_attn.v_experts.50", + "model.layers.38.self_attn.v_experts.51", + "model.layers.38.self_attn.v_experts.52", + "model.layers.38.self_attn.v_experts.53", + "model.layers.38.self_attn.v_experts.54", + "model.layers.38.self_attn.v_experts.55", + "model.layers.38.self_attn.v_experts.56", + "model.layers.38.self_attn.v_experts.57", + "model.layers.38.self_attn.v_experts.58", + "model.layers.38.self_attn.v_experts.59", + "model.layers.38.self_attn.v_experts.60", + "model.layers.38.self_attn.v_experts.61", + "model.layers.38.self_attn.v_experts.62", + "model.layers.38.self_attn.v_experts.63", + "model.layers.38.self_attn.gate_proj", + "model.layers.38.mlp.gate", + "model.layers.38.mlp.shared_experts.gate_proj", + "model.layers.38.mlp.shared_experts.up_proj", + "model.layers.38.mlp.shared_experts.down_proj", + "model.layers.39.self_attn.q_proj", + "model.layers.39.self_attn.k_proj", + "model.layers.39.self_attn.o_proj", + "model.layers.39.self_attn.v_router", + "model.layers.39.self_attn.v_experts.0", + "model.layers.39.self_attn.v_experts.1", + "model.layers.39.self_attn.v_experts.2", + "model.layers.39.self_attn.v_experts.3", + "model.layers.39.self_attn.v_experts.4", + "model.layers.39.self_attn.v_experts.5", + "model.layers.39.self_attn.v_experts.6", + "model.layers.39.self_attn.v_experts.7", + "model.layers.39.self_attn.v_experts.8", + "model.layers.39.self_attn.v_experts.9", + "model.layers.39.self_attn.v_experts.10", + "model.layers.39.self_attn.v_experts.11", + "model.layers.39.self_attn.v_experts.12", + "model.layers.39.self_attn.v_experts.13", + "model.layers.39.self_attn.v_experts.14", + "model.layers.39.self_attn.v_experts.15", + "model.layers.39.self_attn.v_experts.16", + "model.layers.39.self_attn.v_experts.17", + "model.layers.39.self_attn.v_experts.18", + "model.layers.39.self_attn.v_experts.19", + "model.layers.39.self_attn.v_experts.20", + "model.layers.39.self_attn.v_experts.21", + "model.layers.39.self_attn.v_experts.22", + "model.layers.39.self_attn.v_experts.23", + "model.layers.39.self_attn.v_experts.24", + "model.layers.39.self_attn.v_experts.25", + "model.layers.39.self_attn.v_experts.26", + "model.layers.39.self_attn.v_experts.27", + "model.layers.39.self_attn.v_experts.28", + "model.layers.39.self_attn.v_experts.29", + "model.layers.39.self_attn.v_experts.30", + "model.layers.39.self_attn.v_experts.31", + "model.layers.39.self_attn.v_experts.32", + "model.layers.39.self_attn.v_experts.33", + "model.layers.39.self_attn.v_experts.34", + "model.layers.39.self_attn.v_experts.35", + "model.layers.39.self_attn.v_experts.36", + "model.layers.39.self_attn.v_experts.37", + "model.layers.39.self_attn.v_experts.38", + "model.layers.39.self_attn.v_experts.39", + "model.layers.39.self_attn.v_experts.40", + "model.layers.39.self_attn.v_experts.41", + "model.layers.39.self_attn.v_experts.42", + "model.layers.39.self_attn.v_experts.43", + "model.layers.39.self_attn.v_experts.44", + "model.layers.39.self_attn.v_experts.45", + "model.layers.39.self_attn.v_experts.46", + "model.layers.39.self_attn.v_experts.47", + "model.layers.39.self_attn.v_experts.48", + "model.layers.39.self_attn.v_experts.49", + "model.layers.39.self_attn.v_experts.50", + "model.layers.39.self_attn.v_experts.51", + "model.layers.39.self_attn.v_experts.52", + "model.layers.39.self_attn.v_experts.53", + "model.layers.39.self_attn.v_experts.54", + "model.layers.39.self_attn.v_experts.55", + "model.layers.39.self_attn.v_experts.56", + "model.layers.39.self_attn.v_experts.57", + "model.layers.39.self_attn.v_experts.58", + "model.layers.39.self_attn.v_experts.59", + "model.layers.39.self_attn.v_experts.60", + "model.layers.39.self_attn.v_experts.61", + "model.layers.39.self_attn.v_experts.62", + "model.layers.39.self_attn.v_experts.63", + "model.layers.39.self_attn.gate_proj", + "model.layers.39.mlp.gate", + "model.layers.39.mlp.shared_experts.gate_proj", + "model.layers.39.mlp.shared_experts.up_proj", + "model.layers.39.mlp.shared_experts.down_proj", + "model.layers.40.self_attn.q_proj", + "model.layers.40.self_attn.k_proj", + "model.layers.40.self_attn.o_proj", + "model.layers.40.self_attn.v_router", + "model.layers.40.self_attn.v_experts.0", + "model.layers.40.self_attn.v_experts.1", + "model.layers.40.self_attn.v_experts.2", + "model.layers.40.self_attn.v_experts.3", + "model.layers.40.self_attn.v_experts.4", + "model.layers.40.self_attn.v_experts.5", + "model.layers.40.self_attn.v_experts.6", + "model.layers.40.self_attn.v_experts.7", + "model.layers.40.self_attn.v_experts.8", + "model.layers.40.self_attn.v_experts.9", + "model.layers.40.self_attn.v_experts.10", + "model.layers.40.self_attn.v_experts.11", + "model.layers.40.self_attn.v_experts.12", + "model.layers.40.self_attn.v_experts.13", + "model.layers.40.self_attn.v_experts.14", + "model.layers.40.self_attn.v_experts.15", + "model.layers.40.self_attn.v_experts.16", + "model.layers.40.self_attn.v_experts.17", + "model.layers.40.self_attn.v_experts.18", + "model.layers.40.self_attn.v_experts.19", + "model.layers.40.self_attn.v_experts.20", + "model.layers.40.self_attn.v_experts.21", + "model.layers.40.self_attn.v_experts.22", + "model.layers.40.self_attn.v_experts.23", + "model.layers.40.self_attn.v_experts.24", + "model.layers.40.self_attn.v_experts.25", + "model.layers.40.self_attn.v_experts.26", + "model.layers.40.self_attn.v_experts.27", + "model.layers.40.self_attn.v_experts.28", + "model.layers.40.self_attn.v_experts.29", + "model.layers.40.self_attn.v_experts.30", + "model.layers.40.self_attn.v_experts.31", + "model.layers.40.self_attn.v_experts.32", + "model.layers.40.self_attn.v_experts.33", + "model.layers.40.self_attn.v_experts.34", + "model.layers.40.self_attn.v_experts.35", + "model.layers.40.self_attn.v_experts.36", + "model.layers.40.self_attn.v_experts.37", + "model.layers.40.self_attn.v_experts.38", + "model.layers.40.self_attn.v_experts.39", + "model.layers.40.self_attn.v_experts.40", + "model.layers.40.self_attn.v_experts.41", + "model.layers.40.self_attn.v_experts.42", + "model.layers.40.self_attn.v_experts.43", + "model.layers.40.self_attn.v_experts.44", + "model.layers.40.self_attn.v_experts.45", + "model.layers.40.self_attn.v_experts.46", + "model.layers.40.self_attn.v_experts.47", + "model.layers.40.self_attn.v_experts.48", + "model.layers.40.self_attn.v_experts.49", + "model.layers.40.self_attn.v_experts.50", + "model.layers.40.self_attn.v_experts.51", + "model.layers.40.self_attn.v_experts.52", + "model.layers.40.self_attn.v_experts.53", + "model.layers.40.self_attn.v_experts.54", + "model.layers.40.self_attn.v_experts.55", + "model.layers.40.self_attn.v_experts.56", + "model.layers.40.self_attn.v_experts.57", + "model.layers.40.self_attn.v_experts.58", + "model.layers.40.self_attn.v_experts.59", + "model.layers.40.self_attn.v_experts.60", + "model.layers.40.self_attn.v_experts.61", + "model.layers.40.self_attn.v_experts.62", + "model.layers.40.self_attn.v_experts.63", + "model.layers.40.self_attn.gate_proj", + "model.layers.40.mlp.gate", + "model.layers.40.mlp.shared_experts.gate_proj", + "model.layers.40.mlp.shared_experts.up_proj", + "model.layers.40.mlp.shared_experts.down_proj", + "model.layers.41.self_attn.q_proj", + "model.layers.41.self_attn.k_proj", + "model.layers.41.self_attn.o_proj", + "model.layers.41.self_attn.v_router", + "model.layers.41.self_attn.v_experts.0", + "model.layers.41.self_attn.v_experts.1", + "model.layers.41.self_attn.v_experts.2", + "model.layers.41.self_attn.v_experts.3", + "model.layers.41.self_attn.v_experts.4", + "model.layers.41.self_attn.v_experts.5", + "model.layers.41.self_attn.v_experts.6", + "model.layers.41.self_attn.v_experts.7", + "model.layers.41.self_attn.v_experts.8", + "model.layers.41.self_attn.v_experts.9", + "model.layers.41.self_attn.v_experts.10", + "model.layers.41.self_attn.v_experts.11", + "model.layers.41.self_attn.v_experts.12", + "model.layers.41.self_attn.v_experts.13", + "model.layers.41.self_attn.v_experts.14", + "model.layers.41.self_attn.v_experts.15", + "model.layers.41.self_attn.v_experts.16", + "model.layers.41.self_attn.v_experts.17", + "model.layers.41.self_attn.v_experts.18", + "model.layers.41.self_attn.v_experts.19", + "model.layers.41.self_attn.v_experts.20", + "model.layers.41.self_attn.v_experts.21", + "model.layers.41.self_attn.v_experts.22", + "model.layers.41.self_attn.v_experts.23", + "model.layers.41.self_attn.v_experts.24", + "model.layers.41.self_attn.v_experts.25", + "model.layers.41.self_attn.v_experts.26", + "model.layers.41.self_attn.v_experts.27", + "model.layers.41.self_attn.v_experts.28", + "model.layers.41.self_attn.v_experts.29", + "model.layers.41.self_attn.v_experts.30", + "model.layers.41.self_attn.v_experts.31", + "model.layers.41.self_attn.v_experts.32", + "model.layers.41.self_attn.v_experts.33", + "model.layers.41.self_attn.v_experts.34", + "model.layers.41.self_attn.v_experts.35", + "model.layers.41.self_attn.v_experts.36", + "model.layers.41.self_attn.v_experts.37", + "model.layers.41.self_attn.v_experts.38", + "model.layers.41.self_attn.v_experts.39", + "model.layers.41.self_attn.v_experts.40", + "model.layers.41.self_attn.v_experts.41", + "model.layers.41.self_attn.v_experts.42", + "model.layers.41.self_attn.v_experts.43", + "model.layers.41.self_attn.v_experts.44", + "model.layers.41.self_attn.v_experts.45", + "model.layers.41.self_attn.v_experts.46", + "model.layers.41.self_attn.v_experts.47", + "model.layers.41.self_attn.v_experts.48", + "model.layers.41.self_attn.v_experts.49", + "model.layers.41.self_attn.v_experts.50", + "model.layers.41.self_attn.v_experts.51", + "model.layers.41.self_attn.v_experts.52", + "model.layers.41.self_attn.v_experts.53", + "model.layers.41.self_attn.v_experts.54", + "model.layers.41.self_attn.v_experts.55", + "model.layers.41.self_attn.v_experts.56", + "model.layers.41.self_attn.v_experts.57", + "model.layers.41.self_attn.v_experts.58", + "model.layers.41.self_attn.v_experts.59", + "model.layers.41.self_attn.v_experts.60", + "model.layers.41.self_attn.v_experts.61", + "model.layers.41.self_attn.v_experts.62", + "model.layers.41.self_attn.v_experts.63", + "model.layers.41.self_attn.gate_proj", + "model.layers.41.mlp.gate", + "model.layers.41.mlp.shared_experts.gate_proj", + "model.layers.41.mlp.shared_experts.up_proj", + "model.layers.41.mlp.shared_experts.down_proj", + "model.layers.42.self_attn.q_proj", + "model.layers.42.self_attn.k_proj", + "model.layers.42.self_attn.o_proj", + "model.layers.42.self_attn.v_router", + "model.layers.42.self_attn.v_experts.0", + "model.layers.42.self_attn.v_experts.1", + "model.layers.42.self_attn.v_experts.2", + "model.layers.42.self_attn.v_experts.3", + "model.layers.42.self_attn.v_experts.4", + "model.layers.42.self_attn.v_experts.5", + "model.layers.42.self_attn.v_experts.6", + "model.layers.42.self_attn.v_experts.7", + "model.layers.42.self_attn.v_experts.8", + "model.layers.42.self_attn.v_experts.9", + "model.layers.42.self_attn.v_experts.10", + "model.layers.42.self_attn.v_experts.11", + "model.layers.42.self_attn.v_experts.12", + "model.layers.42.self_attn.v_experts.13", + "model.layers.42.self_attn.v_experts.14", + "model.layers.42.self_attn.v_experts.15", + "model.layers.42.self_attn.v_experts.16", + "model.layers.42.self_attn.v_experts.17", + "model.layers.42.self_attn.v_experts.18", + "model.layers.42.self_attn.v_experts.19", + "model.layers.42.self_attn.v_experts.20", + "model.layers.42.self_attn.v_experts.21", + "model.layers.42.self_attn.v_experts.22", + "model.layers.42.self_attn.v_experts.23", + "model.layers.42.self_attn.v_experts.24", + "model.layers.42.self_attn.v_experts.25", + "model.layers.42.self_attn.v_experts.26", + "model.layers.42.self_attn.v_experts.27", + "model.layers.42.self_attn.v_experts.28", + "model.layers.42.self_attn.v_experts.29", + "model.layers.42.self_attn.v_experts.30", + "model.layers.42.self_attn.v_experts.31", + "model.layers.42.self_attn.v_experts.32", + "model.layers.42.self_attn.v_experts.33", + "model.layers.42.self_attn.v_experts.34", + "model.layers.42.self_attn.v_experts.35", + "model.layers.42.self_attn.v_experts.36", + "model.layers.42.self_attn.v_experts.37", + "model.layers.42.self_attn.v_experts.38", + "model.layers.42.self_attn.v_experts.39", + "model.layers.42.self_attn.v_experts.40", + "model.layers.42.self_attn.v_experts.41", + "model.layers.42.self_attn.v_experts.42", + "model.layers.42.self_attn.v_experts.43", + "model.layers.42.self_attn.v_experts.44", + "model.layers.42.self_attn.v_experts.45", + "model.layers.42.self_attn.v_experts.46", + "model.layers.42.self_attn.v_experts.47", + "model.layers.42.self_attn.v_experts.48", + "model.layers.42.self_attn.v_experts.49", + "model.layers.42.self_attn.v_experts.50", + "model.layers.42.self_attn.v_experts.51", + "model.layers.42.self_attn.v_experts.52", + "model.layers.42.self_attn.v_experts.53", + "model.layers.42.self_attn.v_experts.54", + "model.layers.42.self_attn.v_experts.55", + "model.layers.42.self_attn.v_experts.56", + "model.layers.42.self_attn.v_experts.57", + "model.layers.42.self_attn.v_experts.58", + "model.layers.42.self_attn.v_experts.59", + "model.layers.42.self_attn.v_experts.60", + "model.layers.42.self_attn.v_experts.61", + "model.layers.42.self_attn.v_experts.62", + "model.layers.42.self_attn.v_experts.63", + "model.layers.42.self_attn.gate_proj", + "model.layers.42.mlp.gate", + "model.layers.42.mlp.shared_experts.gate_proj", + "model.layers.42.mlp.shared_experts.up_proj", + "model.layers.42.mlp.shared_experts.down_proj", + "model.layers.43.self_attn.q_proj", + "model.layers.43.self_attn.k_proj", + "model.layers.43.self_attn.o_proj", + "model.layers.43.self_attn.v_router", + "model.layers.43.self_attn.v_experts.0", + "model.layers.43.self_attn.v_experts.1", + "model.layers.43.self_attn.v_experts.2", + "model.layers.43.self_attn.v_experts.3", + "model.layers.43.self_attn.v_experts.4", + "model.layers.43.self_attn.v_experts.5", + "model.layers.43.self_attn.v_experts.6", + "model.layers.43.self_attn.v_experts.7", + "model.layers.43.self_attn.v_experts.8", + "model.layers.43.self_attn.v_experts.9", + "model.layers.43.self_attn.v_experts.10", + "model.layers.43.self_attn.v_experts.11", + "model.layers.43.self_attn.v_experts.12", + "model.layers.43.self_attn.v_experts.13", + "model.layers.43.self_attn.v_experts.14", + "model.layers.43.self_attn.v_experts.15", + "model.layers.43.self_attn.v_experts.16", + "model.layers.43.self_attn.v_experts.17", + "model.layers.43.self_attn.v_experts.18", + "model.layers.43.self_attn.v_experts.19", + "model.layers.43.self_attn.v_experts.20", + "model.layers.43.self_attn.v_experts.21", + "model.layers.43.self_attn.v_experts.22", + "model.layers.43.self_attn.v_experts.23", + "model.layers.43.self_attn.v_experts.24", + "model.layers.43.self_attn.v_experts.25", + "model.layers.43.self_attn.v_experts.26", + "model.layers.43.self_attn.v_experts.27", + "model.layers.43.self_attn.v_experts.28", + "model.layers.43.self_attn.v_experts.29", + "model.layers.43.self_attn.v_experts.30", + "model.layers.43.self_attn.v_experts.31", + "model.layers.43.self_attn.v_experts.32", + "model.layers.43.self_attn.v_experts.33", + "model.layers.43.self_attn.v_experts.34", + "model.layers.43.self_attn.v_experts.35", + "model.layers.43.self_attn.v_experts.36", + "model.layers.43.self_attn.v_experts.37", + "model.layers.43.self_attn.v_experts.38", + "model.layers.43.self_attn.v_experts.39", + "model.layers.43.self_attn.v_experts.40", + "model.layers.43.self_attn.v_experts.41", + "model.layers.43.self_attn.v_experts.42", + "model.layers.43.self_attn.v_experts.43", + "model.layers.43.self_attn.v_experts.44", + "model.layers.43.self_attn.v_experts.45", + "model.layers.43.self_attn.v_experts.46", + "model.layers.43.self_attn.v_experts.47", + "model.layers.43.self_attn.v_experts.48", + "model.layers.43.self_attn.v_experts.49", + "model.layers.43.self_attn.v_experts.50", + "model.layers.43.self_attn.v_experts.51", + "model.layers.43.self_attn.v_experts.52", + "model.layers.43.self_attn.v_experts.53", + "model.layers.43.self_attn.v_experts.54", + "model.layers.43.self_attn.v_experts.55", + "model.layers.43.self_attn.v_experts.56", + "model.layers.43.self_attn.v_experts.57", + "model.layers.43.self_attn.v_experts.58", + "model.layers.43.self_attn.v_experts.59", + "model.layers.43.self_attn.v_experts.60", + "model.layers.43.self_attn.v_experts.61", + "model.layers.43.self_attn.v_experts.62", + "model.layers.43.self_attn.v_experts.63", + "model.layers.43.self_attn.gate_proj", + "model.layers.43.mlp.gate", + "model.layers.43.mlp.shared_experts.gate_proj", + "model.layers.43.mlp.shared_experts.up_proj", + "model.layers.43.mlp.shared_experts.down_proj", + "model.layers.44.self_attn.q_proj", + "model.layers.44.self_attn.k_proj", + "model.layers.44.self_attn.o_proj", + "model.layers.44.self_attn.v_router", + "model.layers.44.self_attn.v_experts.0", + "model.layers.44.self_attn.v_experts.1", + "model.layers.44.self_attn.v_experts.2", + "model.layers.44.self_attn.v_experts.3", + "model.layers.44.self_attn.v_experts.4", + "model.layers.44.self_attn.v_experts.5", + "model.layers.44.self_attn.v_experts.6", + "model.layers.44.self_attn.v_experts.7", + "model.layers.44.self_attn.v_experts.8", + "model.layers.44.self_attn.v_experts.9", + "model.layers.44.self_attn.v_experts.10", + "model.layers.44.self_attn.v_experts.11", + "model.layers.44.self_attn.v_experts.12", + "model.layers.44.self_attn.v_experts.13", + "model.layers.44.self_attn.v_experts.14", + "model.layers.44.self_attn.v_experts.15", + "model.layers.44.self_attn.v_experts.16", + "model.layers.44.self_attn.v_experts.17", + "model.layers.44.self_attn.v_experts.18", + "model.layers.44.self_attn.v_experts.19", + "model.layers.44.self_attn.v_experts.20", + "model.layers.44.self_attn.v_experts.21", + "model.layers.44.self_attn.v_experts.22", + "model.layers.44.self_attn.v_experts.23", + "model.layers.44.self_attn.v_experts.24", + "model.layers.44.self_attn.v_experts.25", + "model.layers.44.self_attn.v_experts.26", + "model.layers.44.self_attn.v_experts.27", + "model.layers.44.self_attn.v_experts.28", + "model.layers.44.self_attn.v_experts.29", + "model.layers.44.self_attn.v_experts.30", + "model.layers.44.self_attn.v_experts.31", + "model.layers.44.self_attn.v_experts.32", + "model.layers.44.self_attn.v_experts.33", + "model.layers.44.self_attn.v_experts.34", + "model.layers.44.self_attn.v_experts.35", + "model.layers.44.self_attn.v_experts.36", + "model.layers.44.self_attn.v_experts.37", + "model.layers.44.self_attn.v_experts.38", + "model.layers.44.self_attn.v_experts.39", + "model.layers.44.self_attn.v_experts.40", + "model.layers.44.self_attn.v_experts.41", + "model.layers.44.self_attn.v_experts.42", + "model.layers.44.self_attn.v_experts.43", + "model.layers.44.self_attn.v_experts.44", + "model.layers.44.self_attn.v_experts.45", + "model.layers.44.self_attn.v_experts.46", + "model.layers.44.self_attn.v_experts.47", + "model.layers.44.self_attn.v_experts.48", + "model.layers.44.self_attn.v_experts.49", + "model.layers.44.self_attn.v_experts.50", + "model.layers.44.self_attn.v_experts.51", + "model.layers.44.self_attn.v_experts.52", + "model.layers.44.self_attn.v_experts.53", + "model.layers.44.self_attn.v_experts.54", + "model.layers.44.self_attn.v_experts.55", + "model.layers.44.self_attn.v_experts.56", + "model.layers.44.self_attn.v_experts.57", + "model.layers.44.self_attn.v_experts.58", + "model.layers.44.self_attn.v_experts.59", + "model.layers.44.self_attn.v_experts.60", + "model.layers.44.self_attn.v_experts.61", + "model.layers.44.self_attn.v_experts.62", + "model.layers.44.self_attn.v_experts.63", + "model.layers.44.self_attn.gate_proj", + "model.layers.44.mlp.gate", + "model.layers.44.mlp.shared_experts.gate_proj", + "model.layers.44.mlp.shared_experts.up_proj", + "model.layers.44.mlp.shared_experts.down_proj", + "model.layers.45.self_attn.q_proj", + "model.layers.45.self_attn.k_proj", + "model.layers.45.self_attn.o_proj", + "model.layers.45.self_attn.v_router", + "model.layers.45.self_attn.v_experts.0", + "model.layers.45.self_attn.v_experts.1", + "model.layers.45.self_attn.v_experts.2", + "model.layers.45.self_attn.v_experts.3", + "model.layers.45.self_attn.v_experts.4", + "model.layers.45.self_attn.v_experts.5", + "model.layers.45.self_attn.v_experts.6", + "model.layers.45.self_attn.v_experts.7", + "model.layers.45.self_attn.v_experts.8", + "model.layers.45.self_attn.v_experts.9", + "model.layers.45.self_attn.v_experts.10", + "model.layers.45.self_attn.v_experts.11", + "model.layers.45.self_attn.v_experts.12", + "model.layers.45.self_attn.v_experts.13", + "model.layers.45.self_attn.v_experts.14", + "model.layers.45.self_attn.v_experts.15", + "model.layers.45.self_attn.v_experts.16", + "model.layers.45.self_attn.v_experts.17", + "model.layers.45.self_attn.v_experts.18", + "model.layers.45.self_attn.v_experts.19", + "model.layers.45.self_attn.v_experts.20", + "model.layers.45.self_attn.v_experts.21", + "model.layers.45.self_attn.v_experts.22", + "model.layers.45.self_attn.v_experts.23", + "model.layers.45.self_attn.v_experts.24", + "model.layers.45.self_attn.v_experts.25", + "model.layers.45.self_attn.v_experts.26", + "model.layers.45.self_attn.v_experts.27", + "model.layers.45.self_attn.v_experts.28", + "model.layers.45.self_attn.v_experts.29", + "model.layers.45.self_attn.v_experts.30", + "model.layers.45.self_attn.v_experts.31", + "model.layers.45.self_attn.v_experts.32", + "model.layers.45.self_attn.v_experts.33", + "model.layers.45.self_attn.v_experts.34", + "model.layers.45.self_attn.v_experts.35", + "model.layers.45.self_attn.v_experts.36", + "model.layers.45.self_attn.v_experts.37", + "model.layers.45.self_attn.v_experts.38", + "model.layers.45.self_attn.v_experts.39", + "model.layers.45.self_attn.v_experts.40", + "model.layers.45.self_attn.v_experts.41", + "model.layers.45.self_attn.v_experts.42", + "model.layers.45.self_attn.v_experts.43", + "model.layers.45.self_attn.v_experts.44", + "model.layers.45.self_attn.v_experts.45", + "model.layers.45.self_attn.v_experts.46", + "model.layers.45.self_attn.v_experts.47", + "model.layers.45.self_attn.v_experts.48", + "model.layers.45.self_attn.v_experts.49", + "model.layers.45.self_attn.v_experts.50", + "model.layers.45.self_attn.v_experts.51", + "model.layers.45.self_attn.v_experts.52", + "model.layers.45.self_attn.v_experts.53", + "model.layers.45.self_attn.v_experts.54", + "model.layers.45.self_attn.v_experts.55", + "model.layers.45.self_attn.v_experts.56", + "model.layers.45.self_attn.v_experts.57", + "model.layers.45.self_attn.v_experts.58", + "model.layers.45.self_attn.v_experts.59", + "model.layers.45.self_attn.v_experts.60", + "model.layers.45.self_attn.v_experts.61", + "model.layers.45.self_attn.v_experts.62", + "model.layers.45.self_attn.v_experts.63", + "model.layers.45.self_attn.gate_proj", + "model.layers.45.mlp.gate", + "model.layers.45.mlp.shared_experts.gate_proj", + "model.layers.45.mlp.shared_experts.up_proj", + "model.layers.45.mlp.shared_experts.down_proj", + "model.layers.46.self_attn.q_proj", + "model.layers.46.self_attn.k_proj", + "model.layers.46.self_attn.o_proj", + "model.layers.46.self_attn.v_router", + "model.layers.46.self_attn.v_experts.0", + "model.layers.46.self_attn.v_experts.1", + "model.layers.46.self_attn.v_experts.2", + "model.layers.46.self_attn.v_experts.3", + "model.layers.46.self_attn.v_experts.4", + "model.layers.46.self_attn.v_experts.5", + "model.layers.46.self_attn.v_experts.6", + "model.layers.46.self_attn.v_experts.7", + "model.layers.46.self_attn.v_experts.8", + "model.layers.46.self_attn.v_experts.9", + "model.layers.46.self_attn.v_experts.10", + "model.layers.46.self_attn.v_experts.11", + "model.layers.46.self_attn.v_experts.12", + "model.layers.46.self_attn.v_experts.13", + "model.layers.46.self_attn.v_experts.14", + "model.layers.46.self_attn.v_experts.15", + "model.layers.46.self_attn.v_experts.16", + "model.layers.46.self_attn.v_experts.17", + "model.layers.46.self_attn.v_experts.18", + "model.layers.46.self_attn.v_experts.19", + "model.layers.46.self_attn.v_experts.20", + "model.layers.46.self_attn.v_experts.21", + "model.layers.46.self_attn.v_experts.22", + "model.layers.46.self_attn.v_experts.23", + "model.layers.46.self_attn.v_experts.24", + "model.layers.46.self_attn.v_experts.25", + "model.layers.46.self_attn.v_experts.26", + "model.layers.46.self_attn.v_experts.27", + "model.layers.46.self_attn.v_experts.28", + "model.layers.46.self_attn.v_experts.29", + "model.layers.46.self_attn.v_experts.30", + "model.layers.46.self_attn.v_experts.31", + "model.layers.46.self_attn.v_experts.32", + "model.layers.46.self_attn.v_experts.33", + "model.layers.46.self_attn.v_experts.34", + "model.layers.46.self_attn.v_experts.35", + "model.layers.46.self_attn.v_experts.36", + "model.layers.46.self_attn.v_experts.37", + "model.layers.46.self_attn.v_experts.38", + "model.layers.46.self_attn.v_experts.39", + "model.layers.46.self_attn.v_experts.40", + "model.layers.46.self_attn.v_experts.41", + "model.layers.46.self_attn.v_experts.42", + "model.layers.46.self_attn.v_experts.43", + "model.layers.46.self_attn.v_experts.44", + "model.layers.46.self_attn.v_experts.45", + "model.layers.46.self_attn.v_experts.46", + "model.layers.46.self_attn.v_experts.47", + "model.layers.46.self_attn.v_experts.48", + "model.layers.46.self_attn.v_experts.49", + "model.layers.46.self_attn.v_experts.50", + "model.layers.46.self_attn.v_experts.51", + "model.layers.46.self_attn.v_experts.52", + "model.layers.46.self_attn.v_experts.53", + "model.layers.46.self_attn.v_experts.54", + "model.layers.46.self_attn.v_experts.55", + "model.layers.46.self_attn.v_experts.56", + "model.layers.46.self_attn.v_experts.57", + "model.layers.46.self_attn.v_experts.58", + "model.layers.46.self_attn.v_experts.59", + "model.layers.46.self_attn.v_experts.60", + "model.layers.46.self_attn.v_experts.61", + "model.layers.46.self_attn.v_experts.62", + "model.layers.46.self_attn.v_experts.63", + "model.layers.46.self_attn.gate_proj", + "model.layers.46.mlp.gate", + "model.layers.46.mlp.shared_experts.gate_proj", + "model.layers.46.mlp.shared_experts.up_proj", + "model.layers.46.mlp.shared_experts.down_proj", + "model.layers.47.self_attn.q_proj", + "model.layers.47.self_attn.k_proj", + "model.layers.47.self_attn.o_proj", + "model.layers.47.self_attn.v_router", + "model.layers.47.self_attn.v_experts.0", + "model.layers.47.self_attn.v_experts.1", + "model.layers.47.self_attn.v_experts.2", + "model.layers.47.self_attn.v_experts.3", + "model.layers.47.self_attn.v_experts.4", + "model.layers.47.self_attn.v_experts.5", + "model.layers.47.self_attn.v_experts.6", + "model.layers.47.self_attn.v_experts.7", + "model.layers.47.self_attn.v_experts.8", + "model.layers.47.self_attn.v_experts.9", + "model.layers.47.self_attn.v_experts.10", + "model.layers.47.self_attn.v_experts.11", + "model.layers.47.self_attn.v_experts.12", + "model.layers.47.self_attn.v_experts.13", + "model.layers.47.self_attn.v_experts.14", + "model.layers.47.self_attn.v_experts.15", + "model.layers.47.self_attn.v_experts.16", + "model.layers.47.self_attn.v_experts.17", + "model.layers.47.self_attn.v_experts.18", + "model.layers.47.self_attn.v_experts.19", + "model.layers.47.self_attn.v_experts.20", + "model.layers.47.self_attn.v_experts.21", + "model.layers.47.self_attn.v_experts.22", + "model.layers.47.self_attn.v_experts.23", + "model.layers.47.self_attn.v_experts.24", + "model.layers.47.self_attn.v_experts.25", + "model.layers.47.self_attn.v_experts.26", + "model.layers.47.self_attn.v_experts.27", + "model.layers.47.self_attn.v_experts.28", + "model.layers.47.self_attn.v_experts.29", + "model.layers.47.self_attn.v_experts.30", + "model.layers.47.self_attn.v_experts.31", + "model.layers.47.self_attn.v_experts.32", + "model.layers.47.self_attn.v_experts.33", + "model.layers.47.self_attn.v_experts.34", + "model.layers.47.self_attn.v_experts.35", + "model.layers.47.self_attn.v_experts.36", + "model.layers.47.self_attn.v_experts.37", + "model.layers.47.self_attn.v_experts.38", + "model.layers.47.self_attn.v_experts.39", + "model.layers.47.self_attn.v_experts.40", + "model.layers.47.self_attn.v_experts.41", + "model.layers.47.self_attn.v_experts.42", + "model.layers.47.self_attn.v_experts.43", + "model.layers.47.self_attn.v_experts.44", + "model.layers.47.self_attn.v_experts.45", + "model.layers.47.self_attn.v_experts.46", + "model.layers.47.self_attn.v_experts.47", + "model.layers.47.self_attn.v_experts.48", + "model.layers.47.self_attn.v_experts.49", + "model.layers.47.self_attn.v_experts.50", + "model.layers.47.self_attn.v_experts.51", + "model.layers.47.self_attn.v_experts.52", + "model.layers.47.self_attn.v_experts.53", + "model.layers.47.self_attn.v_experts.54", + "model.layers.47.self_attn.v_experts.55", + "model.layers.47.self_attn.v_experts.56", + "model.layers.47.self_attn.v_experts.57", + "model.layers.47.self_attn.v_experts.58", + "model.layers.47.self_attn.v_experts.59", + "model.layers.47.self_attn.v_experts.60", + "model.layers.47.self_attn.v_experts.61", + "model.layers.47.self_attn.v_experts.62", + "model.layers.47.self_attn.v_experts.63", + "model.layers.47.self_attn.gate_proj", + "model.layers.47.mlp.gate", + "model.layers.47.mlp.shared_experts.gate_proj", + "model.layers.47.mlp.shared_experts.up_proj", + "model.layers.47.mlp.shared_experts.down_proj", + "lm_head" + ], + "kv_cache_scheme": null, + "quant_method": "compressed-tensors", + "quantization_status": "compressed", + "sparsity_config": {}, + "transform_config": {}, + "version": "0.18.1.a20260806" + }, + "query_key_norm": false, + "rms_norm_eps": 1e-06, + "rope_head_dim": 128, + "rope_parameters": { + "rope_theta": 10000000.0, + "rope_type": "default" + }, + "router_aux_loss_coef": 0.001, + "router_scaling_factor": 2.5, + "router_score_func": "sigmoid", + "sliding_window": null, + "tie_word_embeddings": false, + "transformers_version": "5.15.0", + "use_cache": true, + "use_sliding_window": false, + "vocab_size": 250624 +} \ No newline at end of file