Change default onnx opset to 17 when exporting in torch 1.13 or above (…

…#937)

Change default onnx opset to 17  when exprting in torch 1.13 or above.
---------

Co-authored-by: reuvenp <reuvenp@altair-semi.com>

model_compression_toolkit/exporter/model_exporter/pytorch/fakely_quant_onnx_pytorch_exporter.py

@@ -26,8 +26,11 @@
 # ONNX opset version 16 is supported from PyTorch 1.12
 if version.parse(torch.__version__) < version.parse("1.12"):
     OPSET_VERSION = 15
-else:
+elif version.parse("1.12.0") <= version.parse(torch.__version__) < version.parse("1.13.0"):
     OPSET_VERSION = 16
+else:
+    # ONNX opset version 17 is supported from PyTorch 1.13
+    OPSET_VERSION = 17
 
 
 class FakelyQuantONNXPyTorchExporter(BasePyTorchExporter):