feat: 任务书 05,runner 与 tasks 两个模块
runner 用 Docker 跑一条命令并交回结局和资源用量;tasks 读 task.json、准备运行目录、给出正式运行/校验器/轻量检查的运行定义,并从原始产物独立判分。任务目录从参考项目搬到 tasks/circle_packing/,新增 task.json;vocab 的 VERDICTS 删掉"判不了",ProcessResult 的 cancelled 改为"容器从外面被杀了"。 全新上下文审阅后的两处修复:计量 shell 以容器内 root 运行,命令经 setpriv 降权,结果写在容器自己的文件层、结束后 docker cp 取出,并加 no-new-privileges,候选无法伪造计量结果把超时记成正常退出;harness 开跑前删掉旧的 solution.json、metrics.json,旧产物不会被当成本次的值。 Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
@@ -0,0 +1,8 @@
|
||||
{
|
||||
"kind": "circle_packing",
|
||||
"finalMetric": { "id": "sum-of-radii", "direction": "maximize" },
|
||||
"epsilon": 0.01,
|
||||
"knownBest": 2.635983,
|
||||
"run": { "wallSeconds": 120, "cores": 1, "runtime": "runc" },
|
||||
"protectedFiles": ["harness.py", "validate.py", "check-candidate.py", "metrics-spec.json", "solution.json", "evaluate.py", "LICENSE", "SOURCES.md"]
|
||||
}
|
||||
Reference in New Issue
Block a user