Norma_MESBench_1.0 / config /example.yaml
mrfakename's picture
Mirror DongZekai/Norma_MESBench_1.0 from ModelScope
e55dc4e verified
Raw History Blame Contribute Delete
2.26 kB
candidate_model:
api_key: ""
base_url: "https://open.bigmodel.cn/api/coding/paas/v4"
model_name: "glm-4.7"
max_tokens: 64000
temperature: null
top_p: null
top_k: null
enable_thinking: true
heavy_think: false # 新增: 是否开启重度思考模式
h_think_times: 1 # 新增: 重度思考模式下,candidate_model并发调用次数
summary_model: null # 新增: 重度思考模式下,用于总结的模型配置,如果为null则使用candidate_model配置
stream: true
max_retries: 10
concurrency: 4
timeout: 300
judges: [
{
"api_key": "",
"base_url": "https://dashscope.aliyuncs.com/compatible-mode/v1",
"model_name": "kimi-k2-thinking",
"max_tokens": 16000,
"temperature": null,
"top_p": null,
"top_k": null,
"enable_thinking": true,
"stream": true,
"max_retries": 10,
"concurrency": 4,
"timeout": 300
},
{
"api_key": "",
"base_url": "https://dashscope.aliyuncs.com/compatible-mode/v1",
"model_name": "deepseek-r1",
"max_tokens": 16000,
"temperature": null,
"top_p": null,
"top_k": null,
"enable_thinking": true,
"stream": true,
"max_retries": 10,
"concurrency": 10,
"timeout": 300
},
{
"api_key": "",
"base_url": "https://dashscope.aliyuncs.com/compatible-mode/v1",
"model_name": "qwen3-235b-a22b-thinking-2507",
"max_tokens": 16000,
"temperature": null,
"top_p": null,
"top_k": null,
"enable_thinking": true,
"stream": true,
"max_retries": 10,
"concurrency": 10,
"timeout": 300
},
]
datasets_config_path: "./config/test_config.txt"
module_1_path: "./data/1专业技术"
module_2_path: "./data/2通用综合"
module_3_path: "./data/3特色场景"
en_mode: false
result_output_path: "./results/"
weights:
专业技术:
单选: 40
多选: 40
判断: 20
问答: 40
通用综合:
单选: 27
多选: 33
问答: 40
特色场景:
单选: 40
多选: 40
判断: 20
安全权重: 50
质量权重: 50
基础理论权重: 25
合同管理权重: 25
投资控制权重: 25
进度控制权重: 25
医疗权重: 50
机场权重: 50
专业技术权重: 45
通用综合权重: 35
特色场景权重: 20