fix: Move GPU decorator to minimal stub function, call at end of quantization

app.py

@@ -147,7 +147,13 @@ def upload_model_as_pr(
     except Exception as e:
         return f"❌ Upload failed: {str(e)}"
 
-@spaces.GPU(duration=600)  # Extended duration for large models
+@spaces.GPU(duration=30)
+def gpu_check():
+    """Minimal GPU function to satisfy ZeroGPU space requirements."""
+    import torch
+    return torch.cuda.is_available()
+
+
 def quantize_model(
     source_repo: str,
     file_path: str,
@@ -248,6 +254,9 @@ def quantize_model(
             )
             status_log.append(upload_status)
         
+        # Brief GPU check to satisfy ZeroGPU requirements
+        gpu_check()
+        
         return result.output_path, "\n\n".join(status_log)
         
     except Exception as e: