Files
YG_FT/frontend/src/views/system/training-log/trainingLogModel.ts

201 lines
6.7 KiB
TypeScript
Raw Normal View History

import type { EChartsOption } from 'echarts'
import type { FineTuneTask, TrainingLogFile } from '@/types'
import type { FineTuneMetricPoint } from '@/api/modules/fineTune'
export interface TrainingMetricData {
steps: number[]
loss: number[]
gradNorm: number[]
lr: number[]
epoch: number[]
}
export interface TrainingSummary {
epoch: string
trainLoss: string
runtime: string
}
export interface ParsedTrainingLog {
metrics: TrainingMetricData
summary: TrainingSummary
}
const NUMBER_SOURCE = '[-+]?(?:\\d+(?:\\.\\d*)?|\\.\\d+)(?:[eE][-+]?\\d+)?'
function escapeRegExp(value: string) {
return value.replace(/[.*+?^${}()|[\]\\]/g, '\\$&')
}
function extractNumber(source: string, key: string) {
const match = source.match(new RegExp(`['"]?${escapeRegExp(key)}['"]?\\s*(?:=|:)\\s*(${NUMBER_SOURCE})`, 'i'))
return match ? Number(match[1]) : undefined
}
function extractSummaryValue(source: string, key: string) {
const match = source.match(new RegExp(`['"]?${escapeRegExp(key)}['"]?\\s*(?:=|:)\\s*(${NUMBER_SOURCE})`, 'i'))
return match?.[1] || ''
}
/** 根据任务精确选择日志PID 优先,任务名仅作为明确兜底。 */
export function resolveTrainingLogFile(
files: TrainingLogFile[],
task: Pick<FineTuneTask, 'process_id' | 'name'>,
) {
const processId = task.process_id
if (processId != null) {
const pidMatch = files.find((file) => file.pid === processId)
if (pidMatch) return pidMatch
const pidPattern = new RegExp(`(?:^|[^0-9])(?:pid)?${processId}(?:[^0-9]|$)`, 'i')
const filenameMatch = files.find((file) => pidPattern.test(file.file))
if (filenameMatch) return filenameMatch
}
const taskName = task.name.trim()
if (!taskName) return undefined
return files.find((file) => file.name.includes(taskName) || file.file.includes(taskName))
}
/** 解析日志中的逐步指标。字段顺序和常见数值格式均不受限制。 */
export function parseTrainingMetrics(text: string): TrainingMetricData {
const metrics: TrainingMetricData = { steps: [], loss: [], gradNorm: [], lr: [], epoch: [] }
const candidates = text
.split(/\r?\n/)
.flatMap((line) => {
const blocks = line.match(/\{[^{}\r\n]*\}/g)
return blocks?.length ? blocks.map((block) => `${line} ${block}`) : [line]
})
for (const [index, line] of candidates.entries()) {
const loss = extractNumber(line, 'loss')
const gradNorm = extractNumber(line, 'grad_norm')
const learningRate = extractNumber(line, 'learning_rate')
const epoch = extractNumber(line, 'epoch')
if (loss == null && gradNorm == null && learningRate == null) continue
metrics.steps.push(extractNumber(line, 'step') ?? metrics.steps.length + index + 1)
metrics.loss.push(loss ?? Number.NaN)
metrics.gradNorm.push(gradNorm ?? Number.NaN)
metrics.lr.push(learningRate ?? Number.NaN)
metrics.epoch.push(epoch ?? Number.NaN)
}
return metrics
}
export function metricsFromApi(points: FineTuneMetricPoint[]): TrainingMetricData {
const metrics: TrainingMetricData = { steps: [], loss: [], gradNorm: [], lr: [], epoch: [] }
for (const [index, point] of points.entries()) {
const hasMetric = point.loss != null || point.grad_norm != null || point.learning_rate != null
if (!hasMetric) continue
metrics.steps.push(Number(point.step || index + 1))
metrics.loss.push(point.loss == null ? Number.NaN : Number(point.loss))
metrics.gradNorm.push(point.grad_norm == null ? Number.NaN : Number(point.grad_norm))
metrics.lr.push(point.learning_rate == null ? Number.NaN : Number(point.learning_rate))
metrics.epoch.push(point.epoch == null ? Number.NaN : Number(point.epoch))
}
return metrics
}
/** 每次都返回新对象,日志截断或切换时不会残留上一轮汇总。 */
export function parseTrainingSummary(text: string): TrainingSummary {
const emptySummary: TrainingSummary = { epoch: '', trainLoss: '', runtime: '' }
const startMatch = /\*{5}\s*train metrics\s*\*{5}/i.exec(text)
if (!startMatch) return emptySummary
const tail = text.slice(startMatch.index + startMatch[0].length)
const endMatch = /\*{5}\s*train metrics end\s*\*{5}/i.exec(tail)
const body = endMatch ? tail.slice(0, endMatch.index) : tail
return {
epoch: extractSummaryValue(body, 'epoch'),
trainLoss: extractSummaryValue(body, 'train_loss'),
runtime: extractSummaryValue(body, 'train_runtime'),
}
}
export function parseTrainingLog(text: string): ParsedTrainingLog {
return {
metrics: parseTrainingMetrics(text),
summary: parseTrainingSummary(text),
}
}
/** 构建单条训练指标曲线。 */
export function buildMetricChartOption(
label: string,
data: number[],
steps: number[],
color: string,
logScale = false,
): EChartsOption {
const visibleData = data.map((value) => (Number.isFinite(value) ? value : null))
return {
grid: { top: 24, right: 20, bottom: 56, left: 56 },
graphic: visibleData.some((value) => value != null)
? []
: [
{
type: 'text',
left: 'center',
top: 'middle',
style: { text: '暂无训练指标数据', fill: '#94a3b8', fontSize: 13 },
},
],
tooltip: {
trigger: 'axis',
axisPointer: { type: 'cross' },
backgroundColor: 'rgba(15, 23, 42, 0.9)',
borderWidth: 0,
textStyle: { color: '#fff', fontSize: 12 },
},
xAxis: {
type: 'category',
data: steps.map((step, index) => (Number.isFinite(step) ? String(step) : String(index + 1))),
boundaryGap: false,
name: 'Step',
nameTextStyle: { color: '#94a3b8', fontSize: 11 },
axisLine: { lineStyle: { color: '#e2e8f0' } },
axisLabel: { color: '#94a3b8', fontSize: 11 },
splitLine: { show: false },
},
yAxis: {
type: logScale ? 'log' : 'value',
name: label,
nameTextStyle: { color: '#94a3b8', fontSize: 11 },
axisLine: { show: false },
axisTick: { show: false },
axisLabel: { color: '#94a3b8', fontSize: 11 },
splitLine: { lineStyle: { color: '#f1f5f9' } },
},
dataZoom: data.length > 30
? [
{ type: 'inside', start: 0, end: 100 },
{ type: 'slider', height: 16, bottom: 8, borderColor: 'transparent', fillerColor: 'rgba(79,70,229,0.08)', handleStyle: { color: '#4f46e5' } },
]
: [],
series: [
{
name: label,
type: 'line',
data: visibleData,
smooth: true,
symbol: 'none',
lineStyle: { width: 2, color },
areaStyle: {
color: {
type: 'linear',
x: 0,
y: 0,
x2: 0,
y2: 1,
colorStops: [
{ offset: 0, color: `${color}55` },
{ offset: 1, color: `${color}05` },
],
},
},
},
],
}
}