Fix NVIDIA persistence mode and add benchmark results table

This commit is contained in:
Mikhail Chusavitin
2026-04-06 10:47:07 +03:00
parent 6e94216f3b
commit fc5c100a29
6 changed files with 296 additions and 62 deletions

View File

@@ -8,9 +8,12 @@ import (
"os"
"path/filepath"
"sort"
"strconv"
"strings"
"time"
"bee/audit/internal/app"
"bee/audit/internal/platform"
"bee/audit/internal/schema"
)
@@ -161,7 +164,7 @@ func renderPage(page string, opts HandlerOptions) string {
case "benchmark":
pageID = "benchmark"
title = "Benchmark"
body = renderBenchmark()
body = renderBenchmark(opts)
case "tasks":
pageID = "tasks"
title = "Tasks"
@@ -1072,14 +1075,14 @@ func renderValidate(opts HandlerOptions) string {
inv.NVIDIA,
`Select which NVIDIA GPUs to include in Validate. The same selection is used by both NVIDIA GPU cards below and by Validate one by one.`,
`<code>nvidia-smi --query-gpu=index,name,memory.total</code>`,
`<div id="sat-gpu-list"><p style="color:var(--muted);font-size:13px">Loading NVIDIA GPUs…</p></div><div style="display:flex;gap:8px;flex-wrap:wrap;margin-top:8px"><button type="button" class="btn btn-sm btn-secondary" onclick="satSelectAllGPUs()">Select all</button><button type="button" class="btn btn-sm btn-secondary" onclick="satSelectNoGPUs()">Clear</button></div><div id="sat-gpu-selection-note" style="font-size:12px;color:var(--muted);margin-top:8px"></div>`,
)) +
renderSATCard("nvidia", "NVIDIA GPU", "runNvidiaValidateSet('nvidia')", "", renderValidateCardBody(
inv.NVIDIA,
`Runs NVIDIA diagnostics and board inventory checks.`,
`<code>nvidia-smi</code>, <code>dmidecode</code>, <code>dcgmi diag</code>`,
`Runs one GPU at a time on the selected NVIDIA GPUs. Diag level is taken from Validate Profile.`,
`<div id="sat-gpu-list"><p style="color:var(--muted);font-size:13px">Loading NVIDIA GPUs…</p></div><div style="display:flex;gap:8px;flex-wrap:wrap;margin-top:8px"><button type="button" class="btn btn-sm btn-secondary" onclick="satSelectAllGPUs()">Select all</button><button type="button" class="btn btn-sm btn-secondary" onclick="satSelectNoGPUs()">Clear</button></div><div id="sat-gpu-selection-note" style="font-size:12px;color:var(--muted);margin-top:8px"></div>`,
)) +
renderSATCard("nvidia", "NVIDIA GPU", "runNvidiaValidateSet('nvidia')", "", renderValidateCardBody(
inv.NVIDIA,
`Runs NVIDIA diagnostics and board inventory checks.`,
`<code>nvidia-smi</code>, <code>dmidecode</code>, <code>dcgmi diag</code>`,
`Runs one GPU at a time on the selected NVIDIA GPUs. Diag level is taken from Validate Profile.`,
)) +
renderSATCard("nvidia-targeted-stress", "NVIDIA GPU Targeted Stress", "runNvidiaValidateSet('nvidia-targeted-stress')", "", renderValidateCardBody(
inv.NVIDIA,
`Runs a controlled NVIDIA DCGM load in Validate to check stability under moderate stress.`,
@@ -1569,7 +1572,25 @@ func renderSATCard(id, label, runAction, headerActions, body string) string {
// ── Benchmark ─────────────────────────────────────────────────────────────────
func renderBenchmark() string {
type benchmarkHistoryColumn struct {
key string
label string
name string
index int
}
type benchmarkHistoryCell struct {
score float64
present bool
}
type benchmarkHistoryRun struct {
generatedAt time.Time
displayTime string
cells map[string]benchmarkHistoryCell
}
func renderBenchmark(opts HandlerOptions) string {
return `<p style="color:var(--muted);font-size:13px;margin-bottom:16px">Benchmark runs generate a human-readable TXT report and machine-readable result bundle. Tasks continue in the background — view progress in <a href="/tasks">Tasks</a>.</p>
<div class="grid2">
@@ -1618,6 +1639,8 @@ func renderBenchmark() string {
</div>
</div>
` + renderBenchmarkResultsCard(opts.ExportDir) + `
<div id="benchmark-output" style="display:none;margin-top:16px" class="card">
<div class="card-head">Benchmark Output <span id="benchmark-title"></span></div>
<div class="card-body"><div id="benchmark-terminal" class="terminal"></div></div>
@@ -1754,6 +1777,115 @@ benchmarkLoadGPUs();
</script>`
}
func renderBenchmarkResultsCard(exportDir string) string {
columns, runs := loadBenchmarkHistory(exportDir)
if len(runs) == 0 {
return `<div class="card"><div class="card-head">Benchmark Results</div><div class="card-body"><p style="color:var(--muted);font-size:13px">No saved benchmark runs yet.</p></div></div>`
}
var b strings.Builder
b.WriteString(`<div class="card"><div class="card-head">Benchmark Results</div><div class="card-body">`)
b.WriteString(`<p style="color:var(--muted);font-size:13px;margin-bottom:12px">Composite score by saved benchmark run and GPU.</p>`)
b.WriteString(`<div style="overflow-x:auto">`)
b.WriteString(`<table><thead><tr><th>Test</th><th>Time</th>`)
for _, col := range columns {
b.WriteString(`<th>` + html.EscapeString(col.label) + `</th>`)
}
b.WriteString(`</tr></thead><tbody>`)
for i, run := range runs {
b.WriteString(`<tr>`)
b.WriteString(`<td>#` + strconv.Itoa(i+1) + `</td>`)
b.WriteString(`<td>` + html.EscapeString(run.displayTime) + `</td>`)
for _, col := range columns {
cell, ok := run.cells[col.key]
if !ok || !cell.present {
b.WriteString(`<td style="color:var(--muted)">-</td>`)
continue
}
b.WriteString(`<td>` + fmt.Sprintf("%.2f", cell.score) + `</td>`)
}
b.WriteString(`</tr>`)
}
b.WriteString(`</tbody></table></div></div></div>`)
return b.String()
}
func loadBenchmarkHistory(exportDir string) ([]benchmarkHistoryColumn, []benchmarkHistoryRun) {
baseDir := app.DefaultBenchmarkBaseDir
if strings.TrimSpace(exportDir) != "" {
baseDir = filepath.Join(exportDir, "bee-benchmark")
}
paths, err := filepath.Glob(filepath.Join(baseDir, "gpu-benchmark-*", "result.json"))
if err != nil || len(paths) == 0 {
return nil, nil
}
sort.Strings(paths)
columnByKey := make(map[string]benchmarkHistoryColumn)
runs := make([]benchmarkHistoryRun, 0, len(paths))
for _, path := range paths {
raw, err := os.ReadFile(path)
if err != nil {
continue
}
var result platform.NvidiaBenchmarkResult
if err := json.Unmarshal(raw, &result); err != nil {
continue
}
run := benchmarkHistoryRun{
generatedAt: result.GeneratedAt,
displayTime: result.GeneratedAt.Local().Format("2006-01-02 15:04:05"),
cells: make(map[string]benchmarkHistoryCell),
}
for _, gpu := range result.GPUs {
key := benchmarkHistoryColumnKey(gpu.Name, gpu.Index)
columnByKey[key] = benchmarkHistoryColumn{
key: key,
label: benchmarkHistoryColumnLabel(gpu.Name, gpu.Index),
name: strings.TrimSpace(gpu.Name),
index: gpu.Index,
}
run.cells[key] = benchmarkHistoryCell{
score: gpu.Scores.CompositeScore,
present: true,
}
}
runs = append(runs, run)
}
columns := make([]benchmarkHistoryColumn, 0, len(columnByKey))
for _, col := range columnByKey {
columns = append(columns, col)
}
sort.Slice(columns, func(i, j int) bool {
leftName := strings.ToLower(strings.TrimSpace(columns[i].name))
rightName := strings.ToLower(strings.TrimSpace(columns[j].name))
if leftName != rightName {
return leftName < rightName
}
if columns[i].index != columns[j].index {
return columns[i].index < columns[j].index
}
return columns[i].key < columns[j].key
})
sort.Slice(runs, func(i, j int) bool {
return runs[i].generatedAt.After(runs[j].generatedAt)
})
return columns, runs
}
func benchmarkHistoryColumnKey(name string, index int) string {
return strings.TrimSpace(name) + "|" + strconv.Itoa(index)
}
func benchmarkHistoryColumnLabel(name string, index int) string {
name = strings.TrimSpace(name)
if name == "" {
return fmt.Sprintf("GPU %d", index)
}
return fmt.Sprintf("%s / GPU %d", name, index)
}
// ── Burn ──────────────────────────────────────────────────────────────────────
func renderBurn() string {