README: new output directory layout
Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
parent
f9cb34fc17
commit
4aa3385345
11
README.md
11
README.md
@ -122,7 +122,16 @@ evalharness viz show report.json --style excel # 4-sheet 仪表盘
|
||||
evalharness viz show report.json --style errors # 失败样本下钻
|
||||
```
|
||||
|
||||
`--out-dir` 自动产出:`reports/<bench>.report.json`(每样本原始预测、分数明细、token 用量、agent 轨迹)、`viz/<bench>.txt`、`summary.md`(全 bench 一张表)与 `summary.csv`(含 perf 列)。
|
||||
`--out-dir` 产出结构(不传时自动保存到 `evalharness-results/<时间戳>-<模型>/`):
|
||||
|
||||
```
|
||||
<out-dir>/
|
||||
├── summary.xlsx 总表 Excel(Summary/Perf/Categories/Samples 四 sheet,主入口)
|
||||
├── summary.md / .csv 同一张总表的 markdown / csv 版
|
||||
└── <bench>/ 每个 benchmark 一个目录
|
||||
├── report.json 完整报告(每样本预测、分数明细、token、轨迹;重判分用)
|
||||
└── detail.md 该 bench 的 markdown 详情
|
||||
```
|
||||
|
||||
## 5. 缓存与断点
|
||||
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user