feat: 评测任务完成进度收尾与指标维度汇总展示

- platform_store: 评测任务完成时置进度 100% 并落地 completed_time
- eval_runner: 新增指标维度汇总，供雷达图等维度展示
- 前端: 评测详情/列表展示优化、模型管理增强

backend/app/db/platform_store.py

@@ -3361,6 +3361,32 @@ class PlatformStore:
             })
             if new_status == "completed":
                 updates["completed_time"] = utcnow()
+        if new_status == "completed":
+            final_total = int(
+                (result_content or {}).get("sample_count")
+                or task.get("sample_count")
+                or (progress_detail or {}).get("total")
+                or 0
+            )
+            final_completed = int(
+                (result_content or {}).get("completed_count")
+                or task.get("completed_count")
+                or final_total
+            )
+            updates.update({
+                "progress": 100,
+                "progress_detail": {
+                    **progress_detail,
+                    "status": "completed",
+                    "stage": "completed",
+                    "total": final_total,
+                    "completed": max(final_completed, final_total),
+                    "percentage": 100,
+                    "current_index": final_total,
+                    "message": "评测完成",
+                },
+                "completed_time": utcnow(),
+            })
         elif new_status in {"failed", "stopped"}:
             updates.update({
                 "error": job.get("error") or task.get("error") or "",

compute/engines/llama_factory/eval_runner.py

@@ -170,6 +170,30 @@ def _metric_record(score: float | None, sample_count: int, error: str = "", avai
     }
 
 
+def _metric_dimension_summary(metrics: dict[str, Any]) -> list[dict[str, Any]]:
+    labels = {
+        "bleu": "BLEU",
+        "rouge": "ROUGE-L",
+        "cosine": "Cosine 相似度",
+        "exact_match": "精确匹配",
+        "text_similarity": "文本相似度",
+    }
+    result: list[dict[str, Any]] = []
+    for name, item in metrics.items():
+        if not isinstance(item, dict) or item.get("score") is None:
+            continue
+        result.append({
+            "name": labels.get(name, name),
+            "score": float(item.get("score") or 0),
+            "max_score": float(item.get("max_score") or 100),
+            "pass_rate": float(item.get("score") or 0),
+            "sample_count": int(item.get("sample_count") or 0),
+            "available": item.get("available", True),
+            "error": item.get("error", ""),
+        })
+    return result
+
+
 def _rouge_tokens(text: str) -> str:
     text = str(text or "").strip().lower()
     tokens: list[str] = []
@@ -650,7 +674,10 @@ def run_eval(config: dict[str, Any]) -> dict[str, Any]:
             "score": overall_score,
             "max_score": 100,
             "pass_rate": round(passed_count / max(completed, 1) * 100, 1),
-        }]
+            "sample_count": completed,
+            "available": bool(scored),
+            "error": "部分样本未返回可解析评分" if len(scored) < completed else "",
+        }] + _metric_dimension_summary(metrics_result)
         overall_evaluation = f"评测完成：{completed} 样本，{passed_count} 通过，平均 {avg_score}/100 分"
     else:
         passed_count = 0
@@ -661,16 +688,7 @@ def run_eval(config: dict[str, Any]) -> dict[str, Any]:
         ]
         overall_score = round(sum(enabled_scores) / len(enabled_scores), output_precision) if enabled_scores else 0
         overall_score_max = 100
-        dimension_summary = [
+        dimension_summary = _metric_dimension_summary(metrics_result)
-            {
-                "name": name,
-                "score": float(item.get("score") or 0),
-                "max_score": 100,
-                "pass_rate": float(item.get("score") or 0),
-            }
-            for name, item in metrics_result.items()
-            if isinstance(item, dict) and item.get("enabled", True) and item.get("score") is not None
-        ]
         overall_evaluation = f"评测完成：{completed} 样本（未配置 LLM 评委）"
 
     result = {

frontend/src/views/eval/EvalDetailView.vue

@@ -56,7 +56,19 @@ const overallScore = computed(() => formatScore(detail.value?.overall_score, det
 const displayModelName = computed(() => detail.value?.model_name || String(detail.value?.model_id || '-'))
 const displayMetric = computed(() => detail.value?.metric_label || detail.value?.metric || '-')
 const progressDetail = computed(() => detail.value?.progress_detail)
-const progressPercentage = computed(() => Math.max(0, Math.min(100, Math.round(Number(progressDetail.value?.percentage ?? completionRate.value)))))
+const progressTotal = computed(() => Math.max(
+  Number(detail.value?.sample_count || 0),
+  Number(progressDetail.value?.total || 0),
+  Number(detail.value?.samples?.length || 0),
+))
+const progressCompleted = computed(() => detail.value?.status === 'completed'
+  ? progressTotal.value
+  : Math.min(progressTotal.value || Number(detail.value?.completed_count || 0), Number(progressDetail.value?.completed ?? detail.value?.completed_count ?? 0)))
+const progressPercentage = computed(() => detail.value?.status === 'completed'
+  ? 100
+  : detail.value?.status === 'failed' || detail.value?.status === 'stopped'
+    ? Math.max(0, Math.min(100, Math.round(Number(progressDetail.value?.percentage ?? detail.value?.progress ?? completionRate.value))))
+    : Math.max(0, Math.min(100, Math.round(Number(progressDetail.value?.percentage ?? completionRate.value)))))
 const progressStage = computed(() => ({
   dataset: '准备数据集',
   model_loading: '加载模型',
@@ -64,14 +76,32 @@ const progressStage = computed(() => ({
   metrics: '计算指标',
   completed: '评测完成',
   failed: '评测失败',
-}[String(progressDetail.value?.stage || '')] || (detail.value?.status === 'running' ? '任务运行中' : '等待开始')))
+}[detail.value?.status === 'completed' ? 'completed' : String(progressDetail.value?.stage || '')] || (detail.value?.status === 'running' ? '任务运行中' : '等待开始')))
 
-const radarDimensions = computed(() => (detail.value?.dimension_summary || [])
+const radarDimensions = computed(() => {
-  .filter((item) => item.available !== false && Number.isFinite(Number(item.score)))
+  const dimensions = new Map<string, { name: string; value: number }>()
-  .map((item) => ({
+  for (const item of detail.value?.dimension_summary || []) {
-    name: item.name,
+    if (item.available === false || !Number.isFinite(Number(item.score))) continue
-    value: Math.max(0, Math.min(100, Number(item.score) / Math.max(Number(item.max_score) || 100, 1) * 100)),
+    dimensions.set(item.name, {
-  })))
+      name: item.name,
+      value: Math.max(0, Math.min(100, Number(item.score) / Math.max(Number(item.max_score) || 100, 1) * 100)),
+    })
+  }
+  const metricLabels: Record<string, string> = {
+    bleu: 'BLEU',
+    rouge: 'ROUGE-L',
+    cosine: 'Cosine 相似度',
+    exact_match: '精确匹配',
+    text_similarity: '文本相似度',
+  }
+  for (const [key, metric] of Object.entries(detail.value?.basic_metrics || {})) {
+    const score = Number(metric?.score)
+    if (!Number.isFinite(score) || metric?.available === false) continue
+    const name = metricLabels[key] || key
+    if (!dimensions.has(name)) dimensions.set(name, { name, value: Math.max(0, Math.min(100, score)) })
+  }
+  return [...dimensions.values()]
+})
 
 const radarOption = computed<EChartsOption>(() => ({
   tooltip: { trigger: 'item' },
@@ -197,7 +227,7 @@ onUnmounted(stopPolling)
         </div>
         <div class="overview-item">
           <span>评测进度</span>
-          <strong>{{ progressDetail?.completed ?? detail.completed_count }} / {{ progressDetail?.total ?? detail.sample_count }}</strong>
+          <strong>{{ progressCompleted }} / {{ progressTotal }}</strong>
           <el-progress :percentage="progressPercentage" :show-text="false" :stroke-width="5" />
           <small>{{ progressStage }}{{ progressDetail?.message ? ' · ' + progressDetail.message : '' }}</small>
         </div>

frontend/src/views/eval/EvalView.vue

@@ -65,6 +65,20 @@ function displayMetric(row: Partial<EvalTask>) {
   return row.metric_label || row.metric || '-'
 }
 
+function progressPercentage(row: Partial<EvalTask>) {
+  if (row.status === 'completed') return 100
+  return Math.max(0, Math.min(100, Math.round(Number(row.progress_detail?.percentage ?? row.progress ?? 0))))
+}
+
+function progressCompleted(row: Partial<EvalTask>) {
+  if (row.status === 'completed') return row.progress_detail?.total ?? '-'
+  return row.progress_detail?.completed ?? 0
+}
+
+function progressTotal(row: Partial<EvalTask>) {
+  return row.progress_detail?.total ?? '-'
+}
+
 const { start: startPolling, stop: stopPolling } = usePolling(
   async () => {
     await loadEvalList({ silent: true })
@@ -125,18 +139,15 @@ onUnmounted(() => {
                   </el-tooltip>
                 </template>
               </el-table-column>
-              <el-table-column label="评分" prop="score" width="100" align="center" />
+              <el-table-column label="评分" prop="score" width="100" align="center">
+                <template #default="{ row }">
+                  {{ row.score == null ? '-' : `${Number(row.score).toFixed(2)} / 100` }}
+                </template>
+              </el-table-column>
               <el-table-column label="评测进度" width="130" align="center">
                 <template #default="{ row }">
-                  <template v-if="ACTIVE_STATUSES.has(String(row.status || ''))">
+                  <el-progress :percentage="progressPercentage(row)" :stroke-width="6" :show-text="false" />
-                    <el-progress
+                  <small>{{ progressCompleted(row) }} / {{ progressTotal(row) }}</small>
-                      :percentage="Math.max(0, Math.min(100, Math.round(Number(row.progress_detail?.percentage ?? row.progress ?? 0))))"
-                      :stroke-width="6"
-                      :show-text="false"
-                    />
-                    <small>{{ row.progress_detail?.completed ?? 0 }} / {{ row.progress_detail?.total ?? '-' }}</small>
-                  </template>
-                  <span v-else>{{ row.score == null ? '-' : Number(row.score).toFixed(2) + ' / 100' }}</span>
                 </template>
               </el-table-column>
               <el-table-column label="状态" width="100" align="center">

frontend/src/views/model/MergeWeightsView.vue

@@ -56,8 +56,8 @@ async function handleMerge() {
       compute_node_id: form.compute_node_id,
       output_model_name: `${form.model_name}-merged`,
     })
-    ElMessage.success('合并成功')
+    ElMessage.success('合并任务已提交，完成后会自动更新状态')
-    router.push('/model-manage')
+    router.push({ path: '/model-manage', query: { tab: 'trained' } })
   } catch {
     // ignore
   } finally {

frontend/src/views/model/ModelManageView.vue

@@ -1,5 +1,5 @@
 <script setup lang="ts">
-import { computed, onMounted, reactive, ref, watch } from 'vue'
+import { computed, onMounted, onUnmounted, reactive, ref, watch } from 'vue'
 import { useRouter } from 'vue-router'
 import { ElMessage, ElMessageBox } from 'element-plus'
 import DataTablePage from '@/components/DataTablePage.vue'
@@ -20,6 +20,7 @@ import { MODEL_SOURCE_MAP, MODEL_TYPE_MAP, PURPOSE_MAP } from '@/constants'
 import type { ModelItem, TrainedModel } from '@/types'
 import { mergeStatusLabel, mergeStatusType, statusLabel, statusTagType } from '@/utils/status'
 import { useAuthStore } from '@/stores/auth'
+import { usePolling } from '@/composables/usePolling'
 
 const router = useRouter()
 const auth = useAuthStore()
@@ -35,7 +36,7 @@ type TrainedModelRuntime = {
   loaded: boolean
 }
 
-const activeTab = ref<TabKey>('config')
+const activeTab = ref<TabKey>(router.currentRoute.value.query.tab === 'trained' ? 'trained' : 'config')
 const loading = ref(false)
 const configList = ref<ModelItem[]>([])
 const trainedList = ref<TrainedModel[]>([])
@@ -131,19 +132,36 @@ async function handleExport(row: TrainedModel) {
   }
 }
 
-async function loadTrained() {
+async function loadTrained(silent = false) {
-  loading.value = true
+  if (!silent) loading.value = true
   try {
     const res = await getTrainedModels()
     trainedList.value = res?.models || []
   } finally {
-    loading.value = false
+    if (!silent) loading.value = false
   }
 }
 
-function loadData() {
+function hasActiveMerge() {
+  return trainedList.value.some((item) => Boolean(item.merging))
+}
+
+const { start: startTrainedPolling, stop: stopTrainedPolling } = usePolling(
+  async () => {
+    await loadTrained(true)
+    if (!hasActiveMerge()) stopTrainedPolling()
+  },
+  3000,
+  { immediate: false },
+)
+
+async function loadData() {
   if (activeTab.value === 'config') loadConfig()
-  else loadTrained()
+  else {
+    await loadTrained()
+    if (hasActiveMerge()) startTrainedPolling()
+    else stopTrainedPolling()
+  }
 }
 
 async function loadTrainedRuntime(row: TrainedModel, force = false) {
@@ -217,9 +235,16 @@ function handleRefresh() {
   else loadTrained()
 }
 
-watch(activeTab, loadData)
+watch(activeTab, () => {
+  stopTrainedPolling()
+  void loadData()
+})
 
-onMounted(loadData)
+onMounted(() => {
+  void loadData()
+})
+
+onUnmounted(stopTrainedPolling)
 </script>
 
 <template>