Support safe AutoDL instance recovery
Controller CI / test (push) Successful in 1s

This commit is contained in:
2026-10-02 15:23:33 +08:00
parent c7c4850e14
commit b58cc1f5f8
3 changed files with 128 additions and 40 deletions
+2
View File
@@ -59,8 +59,10 @@ ces-controller --help
从 Actions 手动运行 `Manage AutoDL GPU runner`:
- `discover`:列出账号私有镜像和内置公共基础镜像,不产生 GPU 费用;
- `instances`:只读列出当前 Token 账号下的 Pro 实例;
- `provision`:创建 PRO6000、传输模型并注册 GPU Runner;必须填写镜像 UUID,并把确认值
填为 `CREATE`;
- `resume`:对失败后已关机的现有实例继续配置;同样必须填写 `CREATE`;
- `status`:查询实例和 Runner 进程状态;
- `shutdown`:回收训练产物并关闭实例,不释放系统盘。