feat(distill): migrate from go-inference to go-ml Backend

Replace inference.LoadModel() with ml.NewMLXBackend() which wraps
the same Metal model with memory management (SetCacheLimit,
SetMemoryLimit). Replace raw iter.Seq token loop with backend.Chat()
returning Result{Text, Metrics}. Add runtime.GC() between probes
to prevent incremental memory leak.

Reference: go-ml/cmd/cmd_ab.go memory management pattern.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

pkg/lem/distill.go

@@ -8,11 +8,13 @@ import (
 	"log"
 	"os"
 	"path/filepath"
+	"runtime"
 	"strings"
 	"time"
 
 	"forge.lthn.ai/core/go-i18n/reversal"
-	"forge.lthn.ai/core/go-inference"
+	ml "forge.lthn.ai/core/go-ml"
+	"forge.lthn.ai/core/go-mlx"
 )
 
 // DistillProbe is a single input prompt for distillation.
@@ -146,16 +148,34 @@ func RunDistill(args []string) {
 		return
 	}
 
-	// Load model via go-inference.
+	// Set Metal memory limits before loading model.
+	if cacheLimitGB > 0 {
+		mlx.SetCacheLimit(uint64(cacheLimitGB) * 1024 * 1024 * 1024)
+		log.Printf("metal cache limit: %dGB", cacheLimitGB)
+	}
+	if memLimitGB > 0 {
+		mlx.SetMemoryLimit(uint64(memLimitGB) * 1024 * 1024 * 1024)
+		log.Printf("metal memory limit: %dGB", memLimitGB)
+	}
+
+	// Load model via go-ml Backend (wraps go-inference with memory management).
 	log.Printf("loading model: %s", modelCfg.Paths.Base)
-	model, err := inference.LoadModel(modelCfg.Paths.Base)
+	backend, err := ml.NewMLXBackend(modelCfg.Paths.Base)
 	if err != nil {
 		log.Fatalf("load model: %v", err)
 	}
-	defer model.Close()
+	defer backend.Close()
 
-	info := model.Info()
-	log.Printf("model loaded: %s %d-layer", info.Architecture, info.NumLayers)
+	log.Printf("model loaded via %s backend", backend.Name())
+
+	// Build generation options from merged config.
+	genOpts := ml.GenOpts{
+		MaxTokens:     genCfg.MaxTokens,
+		Temperature:   genCfg.Temperature,
+		TopP:          genCfg.TopP,
+		TopK:          genCfg.TopK,
+		RepeatPenalty: genCfg.RepeatPenalty,
+	}
 
 	// Initialise grammar scorer.
 	tok := reversal.NewTokeniser()
@@ -188,27 +208,18 @@ func RunDistill(args []string) {
 
 			// Inference uses bare probe — the model generates from its weights.
 			// Sandwich wrapping is only for the training output format.
-			messages := []inference.Message{
+			messages := []ml.Message{
 				{Role: "user", Content: probe.Prompt},
 			}
 
-			// Generate via native Metal inference.
+			// Generate via go-ml Backend (memory-managed Metal inference).
 			start := time.Now()
-			var sb strings.Builder
-			for token := range model.Chat(ctx, messages,
-				inference.WithMaxTokens(genCfg.MaxTokens),
-				inference.WithTemperature(float32(genCfg.Temperature)),
-				inference.WithTopP(float32(genCfg.TopP)),
-				inference.WithTopK(genCfg.TopK),
-				inference.WithRepeatPenalty(float32(genCfg.RepeatPenalty)),
-			) {
-				sb.WriteString(token.Text)
-			}
-			if err := model.Err(); err != nil {
+			result, err := backend.Chat(ctx, messages, genOpts)
+			if err != nil {
 				fmt.Fprintf(os.Stderr, " → ERROR: %v\n", err)
 				continue
 			}
-			response := sb.String()
+			response := result.Text
 			elapsed := time.Since(start)
 
 			// Quick reject: empty/degenerate.
@@ -222,11 +233,14 @@ func RunDistill(args []string) {
 			delta := ComputeDelta(tok, probe.Prompt, response,
 				aiCfg.Scorer.SycophancyEcho, aiCfg.Scorer.SycophancyUplift)
 
-			met := model.Metrics()
+			tokPerSec := 0.0
+			if result.Metrics != nil {
+				tokPerSec = result.Metrics.DecodeTokensPerSec
+			}
 			fmt.Fprintf(os.Stderr, " → %d chars, g=%.1f up=%+.1f echo=%.2f enr=%+.1f, %.1fs (%.0f tok/s)\n",
 				len(response), grammar.Composite,
 				delta.Uplift, delta.Echo, delta.Enrichment,
-				elapsed.Seconds(), met.DecodeTokensPerSec)
+				elapsed.Seconds(), tokPerSec)
 
 			candidate := &distillCandidate{
 				Response: response,
@@ -266,12 +280,15 @@ func RunDistill(args []string) {
 			fmt.Fprintf(os.Stderr, "  ✗ SKIP %s (best g=%.1f < %.1f)\n",
 				probe.ID, score, *minScore)
 		}
+
+		// Release GPU memory between probes to prevent incremental leak.
+		runtime.GC()
 	}
 
 	duration := time.Since(totalStart)
 
 	fmt.Fprintf(os.Stderr, "\n=== Distillation Complete ===\n")
-	fmt.Fprintf(os.Stderr, "Model:    %s (%s)\n", modelCfg.Name, info.Architecture)
+	fmt.Fprintf(os.Stderr, "Model:    %s (%s)\n", modelCfg.Name, backend.Name())
 	fmt.Fprintf(os.Stderr, "Probes:   %d\n", len(probes))
 	fmt.Fprintf(os.Stderr, "Runs:     %d per probe (%d total generations)\n", *runs, len(probes)**runs)
 	fmt.Fprintf(os.Stderr, "Scorer:   go-i18n/reversal grammar v3, gate >= %.1f\n", *minScore)