Remove unused kwargs from model definitions (#13555)

vllm/model_executor/models/llava_next.py

@@ -12,7 +12,6 @@ from transformers.models.llava_next.modeling_llava_next import (
     get_anyres_image_grid_shape, unpad_image)
 from typing_extensions import NotRequired
 
-from vllm.attention import AttentionMetadata
 from vllm.config import VllmConfig
 from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.sampling_metadata import SamplingMetadata
@@ -508,8 +507,6 @@ class LlavaNextForConditionalGeneration(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,
@@ -571,8 +568,6 @@ class LlavaNextForConditionalGeneration(nn.Module, SupportsMultiModal,
 
         hidden_states = self.language_model.model(input_ids,
                                                   positions,
-                                                  kv_caches,
-                                                  attn_metadata,
                                                   intermediate_tensors,
                                                   inputs_embeds=inputs_embeds)
         return hidden_states