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

test_cases:
  - name: "DeepSeek-V3.2-W8A8-DCP-replicated-indexer"
    model: "vllm-ascend/DeepSeek-V3.2-W8A8"
    envs:
      VLLM_ASCEND_ENABLE_NZ: "1"
      HCCL_OP_EXPANSION_MODE: "AIV"
      OMP_PROC_BIND: "false"
      OMP_NUM_THREADS: "20"
      HCCL_BUFFSIZE: "768"
      PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True"
      VLLM_SERVER_DEV_MODE: "1"
      VLLM_WORKER_MULTIPROC_METHOD: "spawn"
      ASCEND_LAUNCH_BLOCKING: "0"
      ASCEND_ENABLE_USE_FABRIC_MEM: "1"
      VLLM_ASCEND_ENABLE_FLASHCOMM1: "1"
      VLLM_ASCEND_ENABLE_FUSED_MC2: "0"
      VLLM_ASCEND_ENABLE_MLAPO: "0"
      PYTHONHASHSEED: "0"
      ASCEND_A3_ENABLE: "1"
      VLLM_ENGINE_READY_TIMEOUT_S: "10000"
      VLLM_RPC_TIMEOUT: "3600000"
      VLLM_EXECUTE_MODEL_TIMEOUT_SECONDS: "30000"
      TASK_QUEUE_ENABLE: "1"
      CPU_AFFINITY_CONF: "1"
      SERVER_PORT: "DEFAULT_PORT"
    server_cmd:
      - "--port"
      - "$SERVER_PORT"
      - "--seed"
      - "1024"
      - "--max-model-len"
      - "8192"
      - "--max-num-batched-tokens"
      - "1024"
      - "--max-num-seqs"
      - "32"
      - "--trust-remote-code"
      - "--quantization"
      - "ascend"
      - "--data-parallel-size"
      - "1"
      - "--pipeline-parallel-size"
      - "1"
      - "--tensor-parallel-size"
      - "16"
      - "--prefill-context-parallel-size"
      - "1"
      - "--decode-context-parallel-size"
      - "16"
      - "--cp-kv-cache-interleave-size"
      - "1"
      - "--block-size"
      - "128"
      - "--enable-expert-parallel"
      - "--gpu-memory-utilization"
      - "0.95"
      - "--api-server-count"
      - "1"
      - "--safetensors-load-strategy"
      - "prefetch"
      - "--compilation-config"
      - '{"cudagraph_mode": "FULL_DECODE_ONLY", "cudagraph_capture_sizes":[4, 16, 64, 128]}'
      - "--additional-config"
      - '{"enable_dsa_cp": true, "ascend_compilation_config":{"enable_npugraph_ex": true, "enable_static_kernel": false}, "multistream_overlap_shared_expert": true, "enable_mc2_hierarchy_comm": false, "enable_sparse_sfa_c8": true, "enable_sparse_li_c8": true, "enable_cpu_binding": true, "recompute_scheduler_enable": false}'
      - "--speculative-config"
      - '{"num_speculative_tokens": 3, "method": "deepseek_mtp"}'
    test_content: []
    benchmarks:
      acc_gsm8k:
        case_type: accuracy
        dataset_path: vllm-ascend/gsm8k-lite
        request_conf: vllm_api_general_chat
        dataset_conf: gsm8k/gsm8k_gen_0_shot_cot_chat_prompt
        max_out_len: 8192
        batch_size: 32
        baseline: 95
        threshold: 5