[Misc][Model][Refactor] Pass the prefix into Linear layers (#28259)

Signed-off-by: MengqingCao <cmq0113@163.com>

vllm/model_executor/models/gemma2.py

@@ -66,13 +66,22 @@ class Gemma2MLP(nn.Module):
         hidden_act: str,
         hidden_activation: str,
         quant_config: QuantizationConfig | None = None,
+        prefix: str = "",
     ) -> None:
         super().__init__()
         self.gate_up_proj = MergedColumnParallelLinear(
-            hidden_size, [intermediate_size] * 2, bias=False, quant_config=quant_config
+            hidden_size,
+            [intermediate_size] * 2,
+            bias=False,
+            quant_config=quant_config,
+            prefix=f"{prefix}.gate_up_proj",
         )
         self.down_proj = RowParallelLinear(
-            intermediate_size, hidden_size, bias=False, quant_config=quant_config
+            intermediate_size,
+            hidden_size,
+            bias=False,
+            quant_config=quant_config,
+            prefix=f"{prefix}.down_proj",
         )
         if not (hidden_act == hidden_activation == "gelu_pytorch_tanh"):
             raise ValueError(
@@ -134,12 +143,14 @@ class Gemma2Attention(nn.Module):
             self.total_num_kv_heads,
             bias=config.attention_bias,
             quant_config=quant_config,
+            prefix=f"{prefix}.qkv_proj",
         )
         self.o_proj = RowParallelLinear(
             self.total_num_heads * self.head_dim,
             hidden_size,
             bias=config.attention_bias,
             quant_config=quant_config,
+            prefix=f"{prefix}.o_proj",
         )
         self.rotary_emb = get_rope(
             self.head_dim,
@@ -208,6 +219,7 @@ class Gemma2DecoderLayer(nn.Module):
             hidden_act=config.hidden_act,
             hidden_activation=config.hidden_activation,
             quant_config=quant_config,
+            prefix=f"{prefix}.mlp",
         )
         self.input_layernorm = GemmaRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
         self.post_attention_layernorm = GemmaRMSNorm(