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

test_cases:
  - name: "DeepSeek-V3.2-W8A8-TP8-DP2"
    model: "vllm-ascend/DeepSeek-V3.2-W8A8"
    envs:
      HCCL_OP_EXPANSION_MODE: "AIV"
      OMP_PROC_BIND: "false"
      OMP_NUM_THREADS: "1"
      HCCL_BUFFSIZE: "1024"
      VLLM_ASCEND_ENABLE_MLAPO: "1"
      PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True"
      VLLM_ASCEND_ENABLE_FLASHCOMM1: "1"
      VLLM_ENGINE_READY_TIMEOUT_S: "1800"
      SERVER_PORT: "DEFAULT_PORT"
    server_cmd:
      - "--enable-expert-parallel"
      - "--tensor-parallel-size"
      - "8"
      - "--data-parallel-size"
      - "2"
      - "--port"
      - "$SERVER_PORT"
      - "--max-model-len"
      - "8192"
      - "--max-num-batched-tokens"
      - "8192"
      - "--max-num-seqs"
      - "4"
      - "--trust-remote-code"
      - "--quantization"
      - "ascend"
      - "--gpu-memory-utilization"
      - "0.98"
      - "--compilation-config"
      - '{"cudagraph_capture_sizes":[8, 16, 24, 32, 40, 48], "cudagraph_mode":"FULL_DECODE_ONLY"}'
      - "--speculative-config"
      - '{"num_speculative_tokens": 3, "method":"deepseek_mtp"}'
      - "--additional-config"
      - '{"layer_sharding": ["q_b_proj", "o_proj"]}'
      - "--reasoning-parser"
      - "deepseek_v3"
      - "--tokenizer_mode"
      - "deepseek_v32"
    benchmarks:
      acc:
        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: 4096
        batch_size: 8
        baseline: 95
        threshold: 5
      perf_1:
        case_type: performance
        dataset_path: vllm-ascend/GSM8K-in3500-bs400
        request_conf: vllm_api_stream_chat
        dataset_conf: gsm8k/gsm8k_gen_0_shot_cot_str_perf
        num_prompts: 1
        max_out_len: 1500
        batch_size: 1
        request_rate: 11.2
        baseline: 134
        threshold: 0.97
      perf_2:
        case_type: performance
        dataset_path: vllm-ascend/GSM8K-in3500-bs400
        request_conf: vllm_api_stream_chat
        dataset_conf: gsm8k/gsm8k_gen_0_shot_cot_str_perf
        num_prompts: 100
        max_out_len: 1500
        batch_size: 4
        request_rate: 11.2
        baseline: 134
        threshold: 0.97