fix bug of cnn failure (#134)

**Description**
Fix bug of VGG models failed on A100 GPU with batch_size=128.

**Major Revision**
- Add synchronize before delete self.model in _postprocess of superbench/benchmarks/model_benchmarks/pytorch_base.py

**Minor Revision**
- Add if self._gpu_available check 
- Reorder del 

superbench/benchmarks/model_benchmarks/pytorch_base.py

@@ -184,10 +184,12 @@ def _postprocess(self):
             )
             return False
 
-        del self._model
-        del self._optimizer
+        if self._gpu_available:
+            torch.cuda.synchronize()
         del self._target
-
+        del self._optimizer
+        del self._model
+        if self._gpu_available:
             torch.cuda.empty_cache()
 
         return True