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

frontend/src/api/modules/fineTune.ts

@@ -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}`)
 

frontend/src/views/fine-tune/FineTuneListView.vue

@@ -1,7 +1,7 @@
 <script setup lang="ts">
 import { ref, computed, onMounted } from 'vue'
 import { useRouter } from 'vue-router'
-import { ElMessage } from 'element-plus'
+import { ElMessage, ElMessageBox } from 'element-plus'
 import DataTablePage from '@/components/DataTablePage.vue'
 import ModelStatusTag from '@/components/ModelStatusTag.vue'
 import { usePolling } from '@/composables/usePolling'
@@ -10,6 +10,7 @@ import {
   getFineTuneList,
   deleteFineTune,
   stopFineTune,
+  resumeFineTune,
   getFineTuneProgress,
   getFineTune,
 } from '@/api/modules/fineTune'
@@ -87,9 +88,29 @@ async function handleDelete(row: any) {
 }
 
 async function handleStop(row: any) {
+  await ElMessageBox.confirm(
+    '停止后将保留已生成的 checkpoint，可在资源可用时继续训练。是否停止当前任务？',
+    '停止训练',
+    { type: 'warning' },
+  )
   await stopFineTune(row.id)
-  ElMessage.success('训练任务已停止')
-  loadData()
+  ElMessage.success('训练任务已停止，可继续训练')
+  await loadData()
+}
+
+async function handleResume(row: any) {
+  await ElMessageBox.confirm(
+    '系统将使用最新 checkpoint 继续训练，并重新检查基座模型、数据集、算力节点和 GPU 是否可用。是否继续？',
+    '继续训练',
+    { type: 'info' },
+  )
+  try {
+    await resumeFineTune(row.id)
+    ElMessage.success('继续训练已提交')
+    await loadData()
+  } catch {
+    // 请求拦截器已展示后端返回的具体预检原因
+  }
 }
 
 function viewLog(row: any) {
@@ -217,7 +238,7 @@ onMounted(async () => {
     <template #actions="{ row }">
       <div class="action-buttons">
         <el-button
-          v-if="row.status === 'running'"
+          v-if="['syncing', 'queued', 'running'].includes(row.status)"
           type="warning"
           link
           size="small"
@@ -225,6 +246,15 @@ onMounted(async () => {
         >
           <i class="fa fa-stop-circle-o" style="margin-right: 4px" /> 停止
         </el-button>
+        <el-button
+          v-if="row.status === 'stopped' || row.status === 'failed'"
+          type="success"
+          link
+          size="small"
+          @click="handleResume(row)"
+        >
+          <i class="fa fa-play-circle-o" style="margin-right: 4px" /> 继续
+        </el-button>
         <el-button type="primary" link size="small" @click="viewLog(row)">
           <i class="fa fa-file-text-o" style="margin-right: 4px" /> 日志
         </el-button>