As self.assertEqual now compares tensor's dtypes as well as values, there are some number of tests which are unhappy with it. So we temporarily replaced such cases of self.assertEqual with self.assertEqualIgnoreTypes.

However it is temporary measure, as self.assertEqualIgnoreTypes calls need to be reviewed on case-by-case basis and replaced with self.assertEqual calls and proper dtypes or self.assertEqual with exact_dtype=False wherever it is not possible by design.

All this tasks are searchable by: TODO(#38095) after landing of #38102

cc @mruberry