# ==========================================
# ACTUAL TEST CASES
# ==========================================

test_cases:
  - name: "qwen3-32b"
    model: "vllm-ascend/Qwen3-32B-w8a8sc-310-vllm-tp4"
    envs:
      SERVER_PORT: "DEFAULT_PORT"
    server_cmd:
      - "--port"
      - "$SERVER_PORT"
      - "--tensor-parallel-size"
      - "4"
      - "--gpu-memory-utilization"
      - "0.8"
      - "--max-num-seqs"
      - "32"
      - "--dtype"
      - "float16"
      - "--quantization"
      - "ascend"
      - "--max-model-len"
      - "20480"
      - "--no-enable-prefix-caching"
      - "--load_format"
      - "sharded_state"
      - "--compilation-config"
      - '{"cudagraph_mode":"FULL_DECODE_ONLY", "cudagraph_capture_sizes":[8,32]}'
      - "--additional-config"
      - '{"ascend_compilation_config":{"enable_npugraph_ex":false, "fuse_norm_quant": false}}'
      - "--default-chat-template-kwargs"
      - '{"enable_thinking": false}'

    benchmarks:
      perf:
        case_type: performance
        dataset_path: vllm-ascend/GSM8K-in512-bs100-qwen3
        request_conf: vllm_api_stream_chat
        dataset_conf: gsm8k/gsm8k_gen_0_shot_cot_str_perf
        num_prompts: 32
        max_out_len: 512
        batch_size: 8
        request_rate: 0
        baseline: 105.26
        threshold: 0.95
      acc:
        case_type: accuracy
        dataset_path: vllm-ascend/gsm8k
        request_conf: vllm_api_general_chat
        dataset_conf: gsm8k/gsm8k_gen_0_shot_cot_chat_prompt
        max_out_len: 10240
        batch_size: 32
        baseline: 95.07
        threshold: 1
        temperature: 0
        top_p: 0.95