[ux] Switch a warning to debug about a pytorch fallback (#23750)

Signed-off-by: Russell Bryant <rbryant@redhat.com>

vllm/v1/sample/ops/topk_topp_sampler.py

@@ -109,7 +109,7 @@ class TopKTopPSampler(nn.Module):
         # CPU-GPU synchronization while `flashinfer_sample` does.
         if (k is None and p is None) or generators:
             if generators:
-                logger.warning_once("FlashInfer 0.2.3+ does not support "
+                logger.debug_once("FlashInfer 0.2.3+ does not support "
                                   "per-request generators. Falling back to "
                                   "PyTorch-native implementation.")
             return self.forward_native(logits, generators, k, p)