Add heartbeat logging during model loading

Prints a progress line every 15s while each from_pretrained call
blocks, so log streams show the space is alive during cold start.

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

app.py

@@ -46,22 +46,37 @@ print("[startup] all imports done", flush=True)
 
 dtype = torch.bfloat16
 
+
+def _start_heartbeat(label: str) -> threading.Event:
+    done = threading.Event()
+    t0 = time.perf_counter()
+    def _beat():
+        while not done.wait(timeout=15):
+            print(f"[startup] {label} still loading... ({time.perf_counter()-t0:.0f}s)", flush=True)
+    threading.Thread(target=_beat, daemon=True).start()
+    return done
+
+
 _t0_load = time.perf_counter()
 print("[startup] loading transformer from_pretrained (prithivMLmods/Qwen-Image-Edit-Rapid-AIO-V23)...", flush=True)
+_hb = _start_heartbeat("transformer")
 _transformer = QwenImageTransformer2DModel.from_pretrained(
     "prithivMLmods/Qwen-Image-Edit-Rapid-AIO-V23",
     torch_dtype=dtype,
     device_map="cuda",
 )
+_hb.set()
 print(f"[startup] transformer loaded in {time.perf_counter()-_t0_load:.1f}s", flush=True)
 
 _t1_load = time.perf_counter()
 print("[startup] loading pipeline from_pretrained (FireRedTeam/FireRed-Image-Edit-1.1)...", flush=True)
+_hb = _start_heartbeat("pipeline")
 pipe = QwenImageEditPlusPipeline.from_pretrained(
     "FireRedTeam/FireRed-Image-Edit-1.1",
     transformer=_transformer,
     torch_dtype=dtype,
 ).to(device)
+_hb.set()
 print(f"[startup] pipeline loaded in {time.perf_counter()-_t1_load:.1f}s", flush=True)
 
 print("Using default attention processor (FA3 skipped for ZeroGPU GPU-arch compatibility).", flush=True)