Go lem CLI (stdlib + DuckDB) replaces scattered Python scripts: - score: heuristic regex + LLM-as-judge scoring - probe: generate responses then score - compare: diff two score files - status: InfluxDB training/generation progress - export: golden set to training JSONL splits - expand: distributed expansion via API + InfluxDB coordination New scripts from Feb 14 creative session: - scoring_agent.py: ROCm daemon that auto-scores checkpoints - probes.py: 23 binary pass/fail capability probes - convert_adapter.py: MLX to PEFT adapter conversion - score_r1_capability.py: DeepSeek R1 checkpoint scoring - lek_content_scorer.py: 6-dimension ethics content scorer - lem_train_15k.py: InfluxDB-coordinated training script - pipeline.py: DuckDB pipeline (seeds, golden set, expansion) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
237 lines
6.4 KiB
Go
237 lines
6.4 KiB
Go
package main
|
|
|
|
import (
|
|
"flag"
|
|
"fmt"
|
|
"log"
|
|
"os"
|
|
"time"
|
|
)
|
|
|
|
const usage = `Usage: lem <command> [flags]
|
|
|
|
Commands:
|
|
score Score existing response files
|
|
probe Generate responses and score them
|
|
compare Compare two score files
|
|
status Show training and generation progress (InfluxDB + DuckDB)
|
|
export Export golden set to training-format JSONL splits
|
|
expand Generate expansion responses via trained LEM model
|
|
|
|
Set LEM_DB env to default DuckDB path for all commands.
|
|
`
|
|
|
|
func main() {
|
|
if len(os.Args) < 2 {
|
|
fmt.Fprint(os.Stderr, usage)
|
|
os.Exit(1)
|
|
}
|
|
|
|
switch os.Args[1] {
|
|
case "score":
|
|
runScore(os.Args[2:])
|
|
case "probe":
|
|
runProbe(os.Args[2:])
|
|
case "compare":
|
|
runCompare(os.Args[2:])
|
|
case "status":
|
|
runStatus(os.Args[2:])
|
|
case "expand":
|
|
runExpand(os.Args[2:])
|
|
case "export":
|
|
runExport(os.Args[2:])
|
|
default:
|
|
fmt.Fprintf(os.Stderr, "unknown command: %s\n\n%s", os.Args[1], usage)
|
|
os.Exit(1)
|
|
}
|
|
}
|
|
|
|
func runScore(args []string) {
|
|
fs := flag.NewFlagSet("score", flag.ExitOnError)
|
|
|
|
input := fs.String("input", "", "Input JSONL response file (required)")
|
|
suites := fs.String("suites", "all", "Comma-separated suites or 'all'")
|
|
judgeModel := fs.String("judge-model", "mlx-community/gemma-3-27b-it-qat-4bit", "Judge model name")
|
|
judgeURL := fs.String("judge-url", "http://10.69.69.108:8090", "Judge API URL")
|
|
concurrency := fs.Int("concurrency", 4, "Max concurrent judge calls")
|
|
output := fs.String("output", "scores.json", "Output score file path")
|
|
resume := fs.Bool("resume", false, "Resume from existing output, skipping scored IDs")
|
|
|
|
if err := fs.Parse(args); err != nil {
|
|
log.Fatalf("parse flags: %v", err)
|
|
}
|
|
|
|
if *input == "" {
|
|
fmt.Fprintln(os.Stderr, "error: --input is required")
|
|
fs.Usage()
|
|
os.Exit(1)
|
|
}
|
|
|
|
// Read responses.
|
|
responses, err := readResponses(*input)
|
|
if err != nil {
|
|
log.Fatalf("read responses: %v", err)
|
|
}
|
|
log.Printf("loaded %d responses from %s", len(responses), *input)
|
|
|
|
// If resume, load existing scores and filter out already-scored IDs.
|
|
if *resume {
|
|
if _, statErr := os.Stat(*output); statErr == nil {
|
|
existing, readErr := readScorerOutput(*output)
|
|
if readErr != nil {
|
|
log.Fatalf("read existing scores for resume: %v", readErr)
|
|
}
|
|
|
|
// Build set of already-scored IDs.
|
|
scored := make(map[string]bool)
|
|
for _, scores := range existing.PerPrompt {
|
|
for _, ps := range scores {
|
|
scored[ps.ID] = true
|
|
}
|
|
}
|
|
|
|
// Filter out already-scored responses.
|
|
var filtered []Response
|
|
for _, r := range responses {
|
|
if !scored[r.ID] {
|
|
filtered = append(filtered, r)
|
|
}
|
|
}
|
|
log.Printf("resume: skipping %d already-scored, %d remaining",
|
|
len(responses)-len(filtered), len(filtered))
|
|
responses = filtered
|
|
|
|
if len(responses) == 0 {
|
|
log.Println("all responses already scored, nothing to do")
|
|
return
|
|
}
|
|
}
|
|
}
|
|
|
|
// Create client, judge, engine.
|
|
client := NewClient(*judgeURL, *judgeModel)
|
|
client.maxTokens = 512
|
|
judge := NewJudge(client)
|
|
engine := NewEngine(judge, *concurrency, *suites)
|
|
|
|
log.Printf("scoring with %s", engine)
|
|
|
|
// Score all responses.
|
|
perPrompt := engine.ScoreAll(responses)
|
|
|
|
// If resuming, merge with existing scores.
|
|
if *resume {
|
|
if _, statErr := os.Stat(*output); statErr == nil {
|
|
existing, _ := readScorerOutput(*output)
|
|
for model, scores := range existing.PerPrompt {
|
|
perPrompt[model] = append(scores, perPrompt[model]...)
|
|
}
|
|
}
|
|
}
|
|
|
|
// Compute averages and write output.
|
|
averages := computeAverages(perPrompt)
|
|
|
|
scorerOutput := &ScorerOutput{
|
|
Metadata: Metadata{
|
|
JudgeModel: *judgeModel,
|
|
JudgeURL: *judgeURL,
|
|
ScoredAt: time.Now().UTC(),
|
|
ScorerVersion: "1.0.0",
|
|
Suites: engine.SuiteNames(),
|
|
},
|
|
ModelAverages: averages,
|
|
PerPrompt: perPrompt,
|
|
}
|
|
|
|
if err := writeScores(*output, scorerOutput); err != nil {
|
|
log.Fatalf("write scores: %v", err)
|
|
}
|
|
|
|
log.Printf("wrote scores to %s", *output)
|
|
}
|
|
|
|
func runProbe(args []string) {
|
|
fs := flag.NewFlagSet("probe", flag.ExitOnError)
|
|
|
|
model := fs.String("model", "", "Target model name (required)")
|
|
targetURL := fs.String("target-url", "", "Target model API URL (defaults to judge-url)")
|
|
probesFile := fs.String("probes", "", "Custom probes JSONL file (uses built-in content probes if not specified)")
|
|
suites := fs.String("suites", "all", "Comma-separated suites or 'all'")
|
|
judgeModel := fs.String("judge-model", "mlx-community/gemma-3-27b-it-qat-4bit", "Judge model name")
|
|
judgeURL := fs.String("judge-url", "http://10.69.69.108:8090", "Judge API URL")
|
|
concurrency := fs.Int("concurrency", 4, "Max concurrent judge calls")
|
|
output := fs.String("output", "scores.json", "Output score file path")
|
|
|
|
if err := fs.Parse(args); err != nil {
|
|
log.Fatalf("parse flags: %v", err)
|
|
}
|
|
|
|
if *model == "" {
|
|
fmt.Fprintln(os.Stderr, "error: --model is required")
|
|
fs.Usage()
|
|
os.Exit(1)
|
|
}
|
|
|
|
// Default target URL to judge URL.
|
|
if *targetURL == "" {
|
|
*targetURL = *judgeURL
|
|
}
|
|
|
|
// Create clients.
|
|
targetClient := NewClient(*targetURL, *model)
|
|
targetClient.maxTokens = 1024 // Limit probe response length.
|
|
judgeClient := NewClient(*judgeURL, *judgeModel)
|
|
judgeClient.maxTokens = 512 // Judge responses are structured JSON.
|
|
judge := NewJudge(judgeClient)
|
|
engine := NewEngine(judge, *concurrency, *suites)
|
|
prober := NewProber(targetClient, engine)
|
|
|
|
var scorerOutput *ScorerOutput
|
|
var err error
|
|
|
|
if *probesFile != "" {
|
|
// Read custom probes.
|
|
probes, readErr := readResponses(*probesFile)
|
|
if readErr != nil {
|
|
log.Fatalf("read probes: %v", readErr)
|
|
}
|
|
log.Printf("loaded %d custom probes from %s", len(probes), *probesFile)
|
|
|
|
scorerOutput, err = prober.ProbeModel(probes, *model)
|
|
} else {
|
|
log.Printf("using %d built-in content probes", len(contentProbes))
|
|
scorerOutput, err = prober.ProbeContent(*model)
|
|
}
|
|
|
|
if err != nil {
|
|
log.Fatalf("probe: %v", err)
|
|
}
|
|
|
|
if writeErr := writeScores(*output, scorerOutput); writeErr != nil {
|
|
log.Fatalf("write scores: %v", writeErr)
|
|
}
|
|
|
|
log.Printf("wrote scores to %s", *output)
|
|
}
|
|
|
|
func runCompare(args []string) {
|
|
fs := flag.NewFlagSet("compare", flag.ExitOnError)
|
|
|
|
oldFile := fs.String("old", "", "Old score file (required)")
|
|
newFile := fs.String("new", "", "New score file (required)")
|
|
|
|
if err := fs.Parse(args); err != nil {
|
|
log.Fatalf("parse flags: %v", err)
|
|
}
|
|
|
|
if *oldFile == "" || *newFile == "" {
|
|
fmt.Fprintln(os.Stderr, "error: --old and --new are required")
|
|
fs.Usage()
|
|
os.Exit(1)
|
|
}
|
|
|
|
if err := RunCompare(*oldFile, *newFile); err != nil {
|
|
log.Fatalf("compare: %v", err)
|
|
}
|
|
}
|