diff --git a/vllm_mindspore/model_executor/layers/sampler.py b/vllm_mindspore/model_executor/layers/sampler.py
index 172a3faf2059b9740f5ac4a4087200aeda97a0dd..9f01f462c84f4b1482c743d8922cb82a0cd2e570 100644
--- a/vllm_mindspore/model_executor/layers/sampler.py
+++ b/vllm_mindspore/model_executor/layers/sampler.py
@@ -629,7 +629,7 @@ def _multinomial(
                 q[sample_idx : sample_idx + stride]
             )
             sample_idx += stride
-    return probs.div(q).argmax(axis=1).view(-1, num_samples)
+    return probs.div_(q).argmax(dim=1).view(-1, num_samples)
 
 
 def _top_k_top_p_multinomial_with_flashinfer(