{ "base_model": "google/gemma-4-26B-A4B-it", "layers_path": "model.language_model.layers", "num_layers": 30, "layer_class": "Gemma4TextDecoderLayer", "moe_layers": [ { "i": 0, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 1, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 2, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 3, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 4, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 5, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 6, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 7, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 8, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 9, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 10, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 11, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 12, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 13, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 14, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 15, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 16, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 17, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 18, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 19, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 20, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 21, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 22, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 23, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 24, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 25, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 26, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 27, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 28, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null }, { "i": 29, "expert_path": "experts", "expert_pattern": "fused", "num_experts": 128, "gate_path": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "shared_attr": null } ], "dense_layers": [], "expert_attr": "experts", "expert_pattern": "fused", "gate_attr": "router.proj", "router_module": "router", "router_aux_params": [ "per_expert_scale" ], "num_experts_per_layer": 128, "shared_expert_attr": null }