diff --git a/tests/unit_tests/kv_offload/test_offloading_connector.py b/tests/unit_tests/kv_offload/test_offloading_connector.py
index b3eb1bc0e0..a570adf404 100644
--- a/tests/unit_tests/kv_offload/test_offloading_connector.py
+++ b/tests/unit_tests/kv_offload/test_offloading_connector.py
@@ -27,7 +27,7 @@
     init_none_hash,
 )
 from vllm.v1.core.sched.scheduler import Scheduler
-from vllm.v1.kv_cache_interface import KVCacheConfig
+from vllm.v1.kv_cache_interface import FullAttentionSpec, KVCacheConfig, KVCacheGroupSpec
 from vllm.v1.kv_offload.abstract import (
     LoadStoreSpec,
     OffloadingEvent,
@@ -160,8 +160,24 @@ def __init__(self, offloaded_block_size: int, gpu_block_size: int, num_gpu_block
             },
         )
 
+        kv_cache_config = KVCacheConfig(
+            num_blocks=num_gpu_blocks,
+            kv_cache_tensors=[],
+            kv_cache_groups=[
+                KVCacheGroupSpec(
+                    ["layer"],
+                    FullAttentionSpec(
+                        block_size=gpu_block_size,
+                        num_kv_heads=1,
+                        head_size=1,
+                        dtype=torch.float32,
+                    ),
+                )
+            ],
+        )
+
         self.scheduler: Scheduler = create_scheduler(vllm_config, num_blocks=num_gpu_blocks)
-        self.worker_connector = OffloadingConnector(vllm_config, KVConnectorRole.WORKER)
+        self.worker_connector = OffloadingConnector(vllm_config, KVConnectorRole.WORKER, kv_cache_config)
 
         # register worker kv_caches to enable OffloadingWorker creations
         self.worker_connector.register_cross_layers_kv_cache(
diff --git a/vllm_gaudi/v1/kv_offload/worker/cpu_hpu.py b/vllm_gaudi/v1/kv_offload/worker/cpu_hpu.py
index c764305afc..60a94a2d68 100644
--- a/vllm_gaudi/v1/kv_offload/worker/cpu_hpu.py
+++ b/vllm_gaudi/v1/kv_offload/worker/cpu_hpu.py
@@ -360,10 +360,12 @@ def get_handlers(
     attn_backends: dict[str, type[AttentionBackend]],
 ) -> Iterator[tuple[type[LoadStoreSpec], type[LoadStoreSpec], OffloadingHandler]]:
     if not self._handlers:
+        assert len(self.gpu_block_size) == 1
+        gpu_block_size = self.gpu_block_size[0]
         self._handlers = CpuGpuOffloadingHandlers(
             attn_backends=attn_backends,
-            gpu_block_size=self.gpu_block_size,
-            cpu_block_size=self.offloaded_block_size,
+            gpu_block_size=gpu_block_size,
+            cpu_block_size=gpu_block_size * self.block_size_factor,
             num_cpu_blocks=self.num_blocks,
             gpu_caches=kv_caches,
         )