README: agent-env extra dependencies (tau2-bench engine, swebench)
Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
parent
8519e9ae18
commit
97a458d9b8
12
README.md
12
README.md
@ -148,6 +148,18 @@ rep.save('gsm8k.report.json')
|
|||||||
# notebook / async 环境用 await evalharness.arun(...)
|
# notebook / async 环境用 await evalharness.arun(...)
|
||||||
```
|
```
|
||||||
|
|
||||||
|
### Agent 环境的额外依赖
|
||||||
|
|
||||||
|
`tau2_bench` 需要官方引擎(**PyPI 上的 `tau2` 是同名无关项目,别装错**):
|
||||||
|
|
||||||
|
```bash
|
||||||
|
pip install -e /path/to/tau2-bench # 本机源码(es 仓库 tools/ 下有)
|
||||||
|
# 或: pip install git+https://github.com/sierra-research/tau2-bench
|
||||||
|
```
|
||||||
|
|
||||||
|
`swe_bench_verified` 需要 `pip install swebench` + 逐题 Docker 镜像(见 §8)。
|
||||||
|
`bfcl_v3` / `general_fc` 无额外依赖。
|
||||||
|
|
||||||
## 5. 参数速查
|
## 5. 参数速查
|
||||||
|
|
||||||
| 参数 | 作用 |
|
| 参数 | 作用 |
|
||||||
|
|||||||
Loading…
x
Reference in New Issue
Block a user