| 1 | model_name: "Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8" |
| 2 | accuracy_threshold: 0.88 |
| 3 | num_questions: 1319 |
| 4 | num_fewshot: 5 |
| 5 | server_args: "--enforce-eager --max-model-len 8192 --data-parallel-size 2 --enable-expert-parallel --all2all-backend deepep_high_throughput" |
| 6 | env: |
| 7 | VLLM_USE_DEEP_GEMM: "1" |
| 8 | VLLM_USE_DEEP_GEMM_MOE: "1" |