25 lines
895 B
Plaintext
25 lines
895 B
Plaintext
# 复制为 api_test_config.env 后填入真实 key，再 source api_test_config.env
|
||
|
||
# API 认证
|
||
export EVAL_API_KEY="sk-xxxx"
|
||
export EVAL_API_URL="https://api.vectron.meta-stone.com/v1"
|
||
|
||
# 模型名（以 API 服务商的 model id 为准）
|
||
export EVAL_MODEL="DeepSeek/DeepSeek-V4-Flash"
|
||
|
||
# 评测数据集，逗号分隔，想测什么改这里
|
||
# 可用参考：gsm8k,aime24,aime25,aime26,hmmt26,imo_answerbench,competition_math,bbh,drop
|
||
# gpqa_diamond,mmlu_pro,simple_qa,mmlu,cmmlu,arc,hellaswag,trivia_qa,winogrande
|
||
# longbench_v2,openai_mrcr,general_fc,bfcl_v3,bigcodebench,humaneval,live_code_bench
|
||
export EVAL_DATASETS="gsm8k,aime24,arc"
|
||
|
||
# 输出目录标识
|
||
export EVAL_FOLDER_NAME="API-Test"
|
||
|
||
# 评测配置
|
||
export EVAL_CONFIG="config/dpv4-int8_nothinking.yaml"
|
||
export EVAL_BATCH_SIZE=4
|
||
export EVAL_LIMIT="none"
|
||
export EVAL_SEED=42
|
||
export EVAL_THINKING="false"
|