skip indexer_k_cache

vllm/model_executor/models/deepseek_v2.py

@@ -682,13 +682,13 @@ def sparse_attn_indexer(
             quant_block_size,
             scale_fmt,
         )
-    else:
-        ops.indexer_k_cache(
-            k,
-            kv_cache,
-            slot_mapping,
-            scale_fmt,
-        )
+    # else:
+    #     ops.indexer_k_cache(
+    #         k,
+    #         kv_cache,
+    #         slot_mapping,
+    #         scale_fmt,
+    #     )
 
     topk_indices_buffer[: hidden_states.shape[0]] = -1
     if has_prefill: