Fix CI.

Fix tests.

Author: Talmaj

onnx2pytorch/convert/attribute.py

@@ -87,7 +87,10 @@ def extract_attributes(node):
                 kwargs["weight_multiplier"] = extract_attr_values(attr)
         elif attr.name == "auto_pad":
             value = extract_attr_values(attr)
-            if value in ("NOTSET", "VALID", "SAME_UPPER", "SAME_LOWER"):
+            if value == "NOTSET":
+                # NOTSET means explicit padding is used, don't add to kwargs
+                pass
+            elif value in ("VALID", "SAME_UPPER", "SAME_LOWER"):
                 kwargs["auto_pad"] = value
             else:
                 raise NotImplementedError(

tests/onnx2pytorch/operations/test_cast.py

@@ -1,7 +1,6 @@
 import numpy as np
 import torch
 import pytest
-from onnx.mapping import TENSOR_TYPE_TO_NP_TYPE
 
 from onnx2pytorch.operations import Cast