Remove unused kwargs from model definitions (#13555)

vllm/model_executor/models/aria.py

@@ -9,7 +9,6 @@ from transformers import AriaConfig, AriaTextConfig, BatchFeature
 from transformers.models.aria.modeling_aria import AriaCrossAttention
 from transformers.models.aria.processing_aria import AriaProcessor
 
-from vllm.attention import AttentionMetadata
 from vllm.config import CacheConfig, QuantizationConfig, VllmConfig
 from vllm.distributed import get_tensor_model_parallel_rank
 from vllm.model_executor.layers.activation import get_act_fn
@@ -626,8 +625,6 @@ class AriaForConditionalGeneration(nn.Module, SupportsMultiModal):
         self,
         input_ids: torch.Tensor,
         positions: torch.Tensor,
-        kv_caches: List[torch.Tensor],
-        attn_metadata: AttentionMetadata,
         intermediate_tensors: Optional[IntermediateTensors] = None,
         inputs_embeds: Optional[torch.Tensor] = None,
         **kwargs: object,
@@ -643,8 +640,6 @@ class AriaForConditionalGeneration(nn.Module, SupportsMultiModal):
         hidden_states = self.language_model(
             input_ids,
             positions,
-            kv_caches,
-            attn_metadata,
             intermediate_tensors,
             inputs_embeds=inputs_embeds,
         )