[Misc] Add uninitialized params tracking for AutoWeightsLoader (#10327)

Signed-off-by: Isotr0py <2037008807@qq.com>

vllm/model_executor/models/gemma.py

@@ -424,7 +424,8 @@ class GemmaForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
         next_tokens = self.sampler(logits, sampling_metadata)
         return next_tokens
 
-    def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]):
+    def load_weights(self, weights: Iterable[Tuple[str,
+                                                   torch.Tensor]]) -> Set[str]:
         stacked_params_mapping = [
             # (param_name, shard_name, shard_id)
             ("qkv_proj", "q_proj", "q"),
@@ -469,3 +470,4 @@ class GemmaForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
             logger.warning(
                 "Some weights are not initialized from checkpoints: %s",
                 unloaded_params)
+        return loaded_params