diff --git a/vllm_ascend/attention/mla_v1.py b/vllm_ascend/attention/mla_v1.py
index 537cd8e..5d993e0 100644
--- a/vllm_ascend/attention/mla_v1.py
+++ b/vllm_ascend/attention/mla_v1.py
@@ -28,8 +28,6 @@ from vllm_ascend.worker.npu_input_batch import InputBatch
 if TYPE_CHECKING:
     from vllm.v1.core.sched.output import SchedulerOutput
 
-_ALLOWED_NUM_QUERIES_PER_KV = [32, 64, 128]
-
 
 class AscendMLABackend(AttentionBackend):
 
@@ -548,15 +546,6 @@ class AscendMLAImpl(MLAAttentionImpl):
             self.spec_token_num = speculative_config.num_speculative_tokens
             assert self.spec_token_num > 0
 
-        # TODO: support numHeads / numKvHeads < 16 in MLA kernel
-        if self.torchair_graph_enabled:
-            assert self.num_queries_per_kv in _ALLOWED_NUM_QUERIES_PER_KV, \
-                ("The allowed number of queries per kv when enabling both MLA and Graph mode"
-                " only support {32, 64, 128}, Thus this is not supported for DeepSeek-V2-Lite,"
-                " as it only has 16 attention heads. And if you're using DeepSeek-V3 or DeepSeek-R1,"
-                " please make sure after the tensor parallel split, num_heads / num_kv_heads in "
-                "{32, 64, 128}.")
-
     def _v_up_proj_and_o_proj(self, x, enable_multistream_mla: bool = False):
         # Convert from (B, N, L) to (N, B, L)
         x = x.view(-1, self.num_heads, self.kv_lora_rank).transpose(0, 1)