Refactor Worker & InputMetadata (#1843)

vllm/engine/arg_utils.py

@@ -201,9 +201,10 @@ class EngineArgs:
                                    self.dtype, self.seed, self.revision,
                                    self.tokenizer_revision, self.max_model_len,
                                    self.quantization)
-        cache_config = CacheConfig(
-            self.block_size, self.gpu_memory_utilization, self.swap_space,
-            getattr(model_config.hf_config, 'sliding_window', None))
+        cache_config = CacheConfig(self.block_size,
+                                   self.gpu_memory_utilization,
+                                   self.swap_space,
+                                   model_config.get_sliding_window())
         parallel_config = ParallelConfig(self.pipeline_parallel_size,
                                          self.tensor_parallel_size,
                                          self.worker_use_ray,