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

test_cases:
  - name: "Qwen3.5-397B-A17B-w4a8-mtp"
    model: "Eco-Tech/Qwen3.5-397B-A17B-w4a8-mtp"
    envs:
      PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True"
      HCCL_OP_EXPANSION_MODE: "AIV"
      HCCL_BUFFSIZE: "1024"
      OMP_NUM_THREADS: "1"
      TASK_QUEUE_ENABLE: "1"
      SERVER_PORT: "DEFAULT_PORT"
    server_cmd:
      - "--tensor-parallel-size"
      - "8"
      - "--data-parallel-size"
      - "1"
      - "--enable-expert-parallel"
      - "--port"
      - "$SERVER_PORT"
      - "--max-num-seqs"
      - "128"
      - "--quantization"
      - "ascend"
      - "--max-model-len"
      - "133120"
      - "--max-num-batched-tokens"
      - "16384"
      - "--trust-remote-code"
      - "--gpu-memory-utilization"
      - "0.9"
      - "--no-enable-prefix-caching"
      - "--additional-config"
      - '{"enable_cpu_binding":true,"multistream_overlap_shared_expert":true}'
      - "--speculative_config"
      - '{"method": "qwen3_5_mtp", "num_speculative_tokens": 3, "enforce_eager": true}'
      - "--compilation-config"
      - '{"cudagraph_capture_sizes":[1,4,8,12,16,20,24,28,32,36,40,44,48,52,56,60,64,68,72,76,80,84,88,92,96,100,108,112,128,160,172,196,200,212,232,256,260,288,320,360,400], "cudagraph_mode":"FULL_DECODE_ONLY"}'
      - "--allowed-local-media-path"
      - "/"
      - "--mm-processor-cache-gb"
      - "0"
    benchmarks:
      acc_GPQA:
        case_type: accuracy
        dataset_path: vllm-ascend/gpqa
        request_conf: vllm_api_general_chat
        dataset_conf: gpqa/gpqa_gen_0_shot_cot_chat_prompt
        max_out_len: 32768
        batch_size: 32
        baseline: 88.38
        threshold: 5
#      acc_aime2025:
#        case_type: accuracy
#        dataset_path: vllm-ascend/aime2025
#        request_conf: vllm_api_general_chat
#        dataset_conf: aime2025/aime2025_gen_0_shot_chat_prompt
#        max_out_len: 72348
#        batch_size: 32
#        baseline: 93.33
#        threshold: 5
      perf:
        case_type: performance
        dataset_path: vllm-ascend/GSM8K-in3500-bs8000-qwen3
        request_conf: vllm_api_stream_chat
        dataset_conf: gsm8k/gsm8k_gen_0_shot_cot_str_perf
        num_prompts: 256
        max_out_len: 1500
        batch_size: 64
        request_rate: 0
        baseline: 1040
        threshold: 0.97