perf: [Perf script] QWEN3 30B-A3B tensor_parallel_size from 4 to 2 (#1558)

Signed-off-by: Youngeun Kwon <youngeunk@nvidia.com>

Author: youngeunkwon0405

examples/configs/recipes/llm/performance/grpo-qwen3-30ba3b-4n8g.yaml

@@ -31,7 +31,7 @@ policy:
       PYTORCH_CUDA_ALLOC_CONF: expandable_segments:False
   generation:
     vllm_cfg:
-      tensor_parallel_size: 4
+      tensor_parallel_size: 2
 logger:
   log_dir: logs/grpo-qwen3-30ba3b-4n8g
   wandb_enabled: true