feat: 训练任务支持从最新 checkpoint 继续训练

- 后端: 新增 resume/checkpoints 端点,compute 客户端与适配器支持续训
- 计算节点: llama_factory 适配器断点续训支持及测试
- 前端: fine-tune API 封装与列表页续训/断点展示
This commit is contained in:
wuyongtao
2026-08-21 14:48:55 +08:00
parent 0233755859
commit 0b59c1ebee
6 changed files with 276 additions and 7 deletions

View File

@@ -49,6 +49,15 @@ export interface FineTuneGpuStatus {
error?: string
}
export interface FineTuneCheckpoint {
id: string
step: number
name: string
path: string
size_bytes?: number
create_time?: string
}
/** 训练任务列表 */
export const getFineTuneList = () => get<FineTuneTask[]>('/fine-tune')
@@ -89,6 +98,13 @@ export const updateFineTune = (id: string | number, data: Partial<FineTuneTask>)
/** 停止训练任务 */
export const stopFineTune = (id: string | number) => post(`/fine-tune/stop/${id}`)
/** 从最新 checkpoint 继续训练 */
export const resumeFineTune = (id: string | number) => post(`/fine-tune/${id}/resume`)
/** 获取训练断点 */
export const getFineTuneCheckpoints = (id: string | number) =>
get<FineTuneCheckpoint[]>(`/fine-tune/${id}/checkpoints`)
/** 删除训练任务 */
export const deleteFineTune = (id: string | number) => del(`/fine-tune/${id}`)