diff --git a/vllm/v1/worker/hpu_model_runner.py b/vllm/v1/worker/hpu_model_runner.py
index 3f53a42e0b7..7cea2295f41 100644
--- a/vllm/v1/worker/hpu_model_runner.py
+++ b/vllm/v1/worker/hpu_model_runner.py
@@ -2147,6 +2147,7 @@ def warmup_model(self) -> None:
             raise AssertionError("Finished profiling")
         kv_caches = self.kv_caches
         max_blocks = int(kv_caches[0][0].size(0) // self.block_size)
+        self.bucketing_ctx.generate_prompt_buckets()
         self.bucketing_ctx.generate_decode_buckets(max_blocks)
 
         if not htorch.utils.internal.is_lazy(