Spaces:
Running on CPU Upgrade
Running on CPU Upgrade
File size: 3,420 Bytes
efde01b 18e54c4 efde01b 80749c6 efde01b 18e54c4 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 | import fs from 'fs'
import path from 'path'
import type { ModelData, BenchmarkStats, MetricThresholds } from './types'
function pct(val: string): number | null {
const s = val?.trim()
if (!s) return null
const n = parseFloat(s.replace('%', ''))
return isNaN(n) ? null : n / 100
}
function num(val: string): number | null {
const s = val?.trim()
if (!s) return null
const n = parseFloat(s)
return isNaN(n) ? null : n
}
export function loadData(): ModelData[] {
const filePath = path.join(process.cwd(), 'public', 'data.csv')
const text = fs.readFileSync(filePath, 'utf-8')
const lines = text.split('\n').filter(l => l.trim()).slice(2)
return lines
.map(line => {
const cols = line.split(',')
return {
rank: parseInt(cols[0]) || 0,
creator: cols[1]?.trim() || '',
model: cols[2]?.trim() || '',
luc: {
general: pct(cols[3]),
physics: pct(cols[4]),
career: pct(cols[5]),
jd: pct(cols[6]),
avg: pct(cols[7]),
},
rag: {
lcAbs: pct(cols[8]),
lcFact: pct(cols[9]),
hyAbs: pct(cols[10]),
hyFact: pct(cols[11]),
avg: pct(cols[12]),
},
fairness: {
style: num(cols[13]),
con: num(cols[14]),
avg: num(cols[15]),
},
archived: cols[16]?.trim() === 'true',
}
})
.filter(m => m.creator && m.model)
}
export function computeStats(models: ModelData[]): BenchmarkStats {
const active = models.filter(m => !m.archived)
const creators = new Set(models.map(m => m.creator))
const lucScores = active.map(m => m.luc.avg).filter((v): v is number => v !== null)
const ragScores = active.map(m => m.rag.avg).filter((v): v is number => v !== null)
const avgLUC = lucScores.reduce((a, b) => a + b, 0) / lucScores.length
const avgRAG = ragScores.reduce((a, b) => a + b, 0) / ragScores.length
const topModel = active
.filter(m => m.luc.avg !== null)
.sort((a, b) => (b.luc.avg ?? 0) - (a.luc.avg ?? 0))[0]?.model ?? ''
return {
totalModels: models.length,
totalCreators: creators.size,
avgLUC,
avgRAG,
topModel,
lastUpdated: new Date().toLocaleDateString('en-SG', { month: 'short', year: 'numeric' }),
}
}
export function getMaxFairnessAvg(models: ModelData[]): number {
const vals = models.map(m => m.fairness.avg).filter((v): v is number => v !== null)
return Math.max(...vals)
}
function empiricalPercentile(vals: number[], p: number): number {
const sorted = [...vals].sort((a, b) => a - b)
const idx = (p / 100) * (sorted.length - 1)
const lo = Math.floor(idx)
const hi = Math.ceil(idx)
return sorted[lo] + (idx - lo) * (sorted[hi] - sorted[lo])
}
export function computeThresholds(models: ModelData[]): MetricThresholds {
const active = models.filter(m => !m.archived)
const luc = active.map(m => m.luc.avg).filter((v): v is number => v !== null)
const rag = active.map(m => m.rag.avg).filter((v): v is number => v !== null)
const fair = active.map(m => m.fairness.avg).filter((v): v is number => v !== null)
return {
luc: { p33: empiricalPercentile(luc, 33), p67: empiricalPercentile(luc, 67) },
rag: { p33: empiricalPercentile(rag, 33), p67: empiricalPercentile(rag, 67) },
fairness: { p33: empiricalPercentile(fair, 33), p67: empiricalPercentile(fair, 67) },
}
}
|