Diagnose single-thread CPU model loading

app.py

@@ -304,8 +304,8 @@ def _run_cpu_smoke(model_filter):
         "name": model["name"],
         "engine": "llama-cli",
         "gpu_layers": 0,
-        "threads": _parse_cpu_count() or 8,
-        "no_mmap": True,
+        "threads": 1,
+        "no_mmap": False,
         "phases": {},
     }
 
@@ -329,7 +329,14 @@ def _run_cpu_smoke(model_filter):
         result["phases"]["download"] = {"ok": False, "elapsed_sec": round(time.time() - t0, 2), "error": str(exc)}
         return result
 
-    command = [cli, "-m", path, "-t", str(result["threads"]), "-ngl", "0", "-c", "256", "-n", "1", "-p", "Say hi.", "--no-display-prompt", "--no-warmup", "--no-mmap", "-fa", "0"]
+    help_t0 = time.time()
+    try:
+        help_run = subprocess.run([cli, "--help"], capture_output=True, text=True, timeout=15, env=env)
+        result["phases"]["help"] = {"ok": help_run.returncode == 0, "elapsed_sec": round(time.time() - help_t0, 2), "returncode": help_run.returncode}
+    except Exception as exc:
+        result["phases"]["help"] = {"ok": False, "elapsed_sec": round(time.time() - help_t0, 2), "error": str(exc)}
+
+    command = [cli, "-m", path, "-t", "1", "-ngl", "0", "-c", "256", "-n", "1", "-p", "Say hi.", "--no-display-prompt", "--no-warmup", "-fa", "0"]
     t0 = time.time()
     try:
         run = subprocess.run(command, capture_output=True, text=True, timeout=180, env=env)