# ==========================================
# Shared Configurations
# ==========================================

_envs: &envs
  VLLM_USE_MODELSCOPE: "true"
  HCCL_OP_EXPANSION_MODE: "AIV"
  HCCL_BUFFSIZE: "1024"
  OMP_PROC_BIND: "false"
  OMP_NUM_THREADS: "1"
  PYTORCH_NPU_ALLOC_CONF: "expandable_segments:True"
  SERVER_PORT: "DEFAULT_PORT"

_server_cmd: &server_cmd
  - "--tensor-parallel-size"
  - "16"
  - "--enable-expert-parallel"
  - "--enable-ep-weight-filter"
  - "--tool-call-parser"
  - "hy_v3"
  - "--reasoning-parser"
  - "hy_v3"
  - "--enable-auto-tool-choice"
  - "--max-model-len"
  - "32768"
  - "--max-num-seqs"
  - "8"
  - "--port"
  - "$SERVER_PORT"
  - "--speculative-config"
  - '{"method": "mtp", "num_speculative_tokens": 1}'
  - "--trust-remote-code"
  - "--gpu-memory-utilization"
  - "0.9"

_benchmarks: &benchmarks
  acc_gsm8k:
    case_type: accuracy
    dataset_path: vllm-ascend/gsm8k-lite
    request_conf: vllm_api_general_chat
    dataset_conf: gsm8k/gsm8k_gen_4_shot_cot_chat_prompt
    max_out_len: 4096
    batch_size: 8
    baseline: 93.07
    threshold: 10

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

test_cases:
  - name: "Hy3-preview-TP16-EP-MTP"
    model: "Tencent-Hunyuan/Hy3-preview"
    envs:
      <<: *envs
    server_cmd: *server_cmd
    benchmarks:
      <<: *benchmarks