CodeBench: --rounds N toistaa testiajot 1-10 kertaa

Kierrosyhteenveto näyttää mediaanin, min/max ja pass-raten per kierros.
Käyttö: node benchmark.mjs --models qwen3:14b --scenarios all --rounds 3

kipina-codebench/benchmark.mjs

@@ -35,6 +35,7 @@ const RESULTS_DIR = join(__dirname, 'results');
 const THINK_MODE = args.includes('--think');
 const COMPACT_MODE = args.includes('--compact');
 const LANG = arg('lang', 'python');  // python | rust
+const ROUNDS = parseInt(arg('rounds', '1'));  // 1-10 toistoa
 const MAX_FIX_ROUNDS = 2;
 
 // === Promptien lataus tiedostoista ===
@@ -404,6 +405,7 @@ async function main() {
 
     const scenarios = SCENARIO_FILTER === 'all' ? SCENARIOS : [SCENARIOS[0]];
     console.log(`Skenaariot (${scenarios.length}): ${scenarios.map(s => s.id).join(', ')}`);
+    if (ROUNDS > 1) console.log(`Toistoja: ${ROUNDS}`);
     console.log(`Tulokset: ${OUTPUT_DIR}/`);
     console.log('');
 
@@ -413,10 +415,14 @@ async function main() {
 
     const results = [];
 
+    for (let round = 1; round <= ROUNDS; round++) {
+    if (ROUNDS > 1) console.log(`\n╔═══ Kierros ${round}/${ROUNDS} ═══╗`);
     for (const model of models) {
         for (const scenario of scenarios) {
-            console.log(`\n━━━ ${model} × ${scenario.id} ━━━`);
+            const roundLabel = ROUNDS > 1 ? ` [${round}/${ROUNDS}]` : '';
+            console.log(`\n━━━ ${model} × ${scenario.id}${roundLabel} ━━━`);
             const r = await runPipeline(model, scenario);
+            if (ROUNDS > 1) r.round = round;
             results.push(r);
 
             const status = r.error ? `✗ ${r.error}` :
@@ -426,6 +432,7 @@ async function main() {
             console.log(`    → ${status} | ${r.stars} ${r.score}p | ${(r.totalDurationMs/1000).toFixed(1)}s | ${r.totalTokens} tok | ${r.avgTokPerSec.toFixed(1)} tok/s${ctxInfo}`);
         }
     }
+    } // rounds
 
     // === Tulostaulu ===
     console.log('\n\n╔══════════════════════════════════════════════════════════════════════════════════════════════════╗');
@@ -540,6 +547,28 @@ async function main() {
     const avgScore = results.length > 0 ? Math.round(results.reduce((s, r) => s + r.score, 0) / results.length) : 0;
     const totalTime = results.reduce((s, r) => s + r.totalDurationMs, 0);
     console.log(`\n${starsForScore(avgScore)} Keskiarvo: ${avgScore}p | ✓ PASS: ${passed.length} | ◐ PARTIAL: ${partial.length} | ✗ FAIL: ${failed.length} | Yhteensä: ${results.length} | Kokonaisaika: ${(totalTime/1000/60).toFixed(1)} min`);
+
+    // === Kierrosyhteenveto (kun rounds > 1) ===
+    if (ROUNDS > 1) {
+        console.log('\n\n╔══════════════════════════════════════════════╗');
+        console.log('║       KIERROSYHTEENVETO (mediaani)            ║');
+        console.log('╚══════════════════════════════════════════════╝\n');
+        const median = arr => { const s = [...arr].sort((a,b) => a-b); const m = Math.floor(s.length/2); return s.length % 2 ? s[m] : Math.round((s[m-1]+s[m])/2); };
+
+        for (const model of modelNames) {
+            const mrs = results.filter(r => r.model === model);
+            for (const sid of scenarioIds) {
+                const runs = mrs.filter(r => r.scenario === sid);
+                if (runs.length === 0) continue;
+                const scores = runs.map(r => r.score);
+                const med = median(scores);
+                const min = Math.min(...scores);
+                const max = Math.max(...scores);
+                const passRates = runs.map(r => r.testsTotal > 0 ? Math.round(r.testsPassed/r.testsTotal*100) : 0);
+                console.log(`${model.padEnd(30)} ${sid.padEnd(8)} ${starsForScore(med)} med:${med}p  min:${min} max:${max}  pass:[${passRates.join(',')}]%`);
+            }
+        }
+    }
 }
 
 main().catch(e => { console.error(e); process.exit(1); });