[BugFix][CPU] Fix TorchSDPABackendImpl doesn't have use_irope (vllm-project#21200)

Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: Himanshu Jaju <hj@mistral.ai>

Author: LucasWilkinson

vllm/v1/worker/gpu_model_runner.py

@@ -2668,7 +2668,8 @@ def get_kv_cache_spec(self) -> dict[str, KVCacheSpec]:
             # TODO: Support other attention modules, e.g., cross-attention
             if attn_module.attn_type == AttentionType.DECODER:
                 use_local_attention = (self.attention_chunk_size is not None
-                                       and attn_module.impl.use_irope)
+                                       and getattr(attn_module.impl,
+                                                   "use_irope", False))
                 if attn_module.sliding_window is not None:
                     kv_cache_spec[layer_name] = SlidingWindowSpec(
                         block_size=block_size,