feat: 数据处理向导新增生成控制选项

提取 GenerationOptionsPanel 组件统一管理生成模型、温度、最大长度、JSON 模式与质量过滤配置,StructuredProcessOptions 与 UnstructuredProcessOptions 继承 GenerationControlOptions,CreateView、TaskSetupStep 及各子步骤同步接入,详情/列表小幅调整,回归脚本扩充生成选项断言。
This commit is contained in:
caoxiaozhu
2026-07-13 11:47:33 +08:00
parent b64d730616
commit 5c469a1778
12 changed files with 744 additions and 30 deletions

View File

@@ -18,7 +18,6 @@ interface ProcessDetail {
status: string
description: string
processType: string
dataType: string
sourceDataset: string
outputDataset?: string
outputDatasetId?: number
@@ -75,8 +74,7 @@ const detailMap: Record<string, ProcessDetail> = {
name: '客服问答数据清洗',
status: 'completed',
description: '清洗客服对话中的无效记录、重复问答和格式异常内容,生成可用于模型训练的标准数据集。',
processType: '去重清洗',
dataType: '结构化数据',
processType: '结构化数据',
sourceDataset: '客服对话原始集',
outputDataset: '客服对话清洗集',
outputDatasetId: 7,
@@ -101,7 +99,7 @@ const detailMap: Record<string, ProcessDetail> = {
},
'492015': {
id: '492015', name: '指令微调数据构造', status: 'running',
description: '从通用语料中构造指令微调训练样本。', processType: '指令构造', dataType: '非结构化数据',
description: '从通用语料中构造指令微调训练样本。', processType: '非结构化数据',
sourceDataset: '通用语料库', outputDataset: 'SFT 指令集', outputDatasetId: 8, creator: '管理员',
createTime: '2026-07-09 09:10:00', startTime: '2026-07-09 09:10:21', duration: '处理中', progress: 68,
inputCount: 18500, outputCount: 8568, filteredCount: 425, duplicateCount: 192, errorCount: 8,
@@ -112,7 +110,7 @@ const detailMap: Record<string, ProcessDetail> = {
},
'731948': {
id: '731948', name: '敏感信息脱敏处理', status: 'pending', description: '识别并脱敏用户反馈数据中的敏感字段。',
processType: '脱敏处理', dataType: '结构化数据', sourceDataset: '用户反馈数据', outputDataset: '用户反馈脱敏集',
processType: '结构化数据', sourceDataset: '用户反馈数据', outputDataset: '用户反馈脱敏集',
outputDatasetId: 9, creator: '管理员', createTime: '2026-07-09 16:45:00', duration: '等待执行', progress: 0,
inputCount: 9340, outputCount: 0, filteredCount: 0, duplicateCount: 0, errorCount: 0,
config: [
@@ -122,7 +120,7 @@ const detailMap: Record<string, ProcessDetail> = {
},
'582012': {
id: '582012', name: '多轮对话拼接', status: 'failed', description: '将单轮问答按会话标识拼接为多轮对话数据。',
processType: '格式转换', dataType: '结构化数据', sourceDataset: '单轮问答集', creator: '管理员',
processType: '结构化数据', sourceDataset: '单轮问答集', creator: '管理员',
createTime: '2026-07-10 08:30:00', startTime: '2026-07-10 08:30:16', completeTime: '2026-07-10 08:31:04',
duration: '48 秒', progress: 37, inputCount: 7520, outputCount: 2780, filteredCount: 24, duplicateCount: 0, errorCount: 1,
config: [
@@ -186,7 +184,6 @@ function resetPage() {
<dl class="heading-meta">
<div><dt>任务 ID</dt><dd>{{ detail.id }}</dd></div>
<div><dt>处理类型</dt><dd>{{ detail.processType }}</dd></div>
<div><dt>数据类型</dt><dd>{{ detail.dataType }}</dd></div>
<div><dt>创建人</dt><dd>{{ detail.creator }}</dd></div>
</dl>
</div>